Flock Safety Cuts Data Storage to Seven Days as Abuse Audits Terminate Officers

Flock Safety将数据存储缩短至7天 滥用审计致涉事警员被解职 Garrett Langley认为，公共安全技术必须执行严格的7天数据保留期和自动滥用审计，以维持市政信任并防止大规模监控。 Flock Safety首席执行官Garrett Langley针对其公司的车牌识别系统、无人机及执法软件引发的激烈争议和公共政策辩论作出了回应。他阐述了Flock默认的7天数据保留政策，明确表示坚决不开发人脸识别或预测性警务工具，并介绍了新部署的自动化审计功能，该功能此前已查出佐治亚州警员滥用系统的行为。尽管在多个城市遭遇了摄像头被破坏和合同被取消的情况，Langley仍坚称，兼顾平衡的防护机制与人工监督对于在保障公共安全的同时维护公民自由至关重要。 Flock Safety的使命是利用车牌识别（LPR）技术消除犯罪，目前业务已拓展至6000多个城市，协助侦破了超过100万起案件。 (0:00) Flock将其默认的数据保留期从30天缩短至7天，在解决约90%犯罪案件的同时，最大限度地降低了隐私风险。 (5:48) Flock在其整个网络中刻意排除了侵入性功能，例如人脸识别、实时视频流、车厢内部查看或基于个人的搜索。 (11:02) 蓄意破坏市政安全摄像头属于重罪，会直接导致安珀警报（Amber Alerts）等关键执法工具失效。 (15:33) 为防止警察不当行为和未经授权的追踪，Flock已向所有客户强制推行其能够发出红旗预警的自动化“审计协助”（Audit Assistance）工具。 (17:28)

Flock Safety CEO Garrett Langley contends that automated license plate readers and rapid-response drones provide vital public safety gains, arguing that vendor-enforced data limits, automated audit software, and human oversight are necessary to curb police misuse and address surveillance fears. Architectural Guardrails and Data Retention Policies (5:48) Operating across roughly 6,000 U.S. municipalities, Flock reports assisting in more than 1,000,000 solved crimes and locating over 10,000 missing persons, including 1,125 individuals found and 20,000 stolen vehicles tracked in July 2026 alone. Langley contrasts Flock's license plate reading (LPR) architecture with competing surveillance vendors, noting that Flock explicitly excludes facial recognition, live video streams, interior vehicle cabin views, and person-based searches. The system instead indexes vehicle metadata such as make, model, altered plates, bumper stickers, and physical damage. To balance operational utility with civil liberties, Flock recently lowered its default data retention period from 30 days to 7 days. Langley states that company data shows approximately 90% of crimes resolved via Flock are solved within that first week. The remaining 10%—such as delayed homicide investigations—require longer windows, which are shaped by divergent state laws ranging from New Hampshire's 3-minute limit and New York's 21-day cap to Washington's 30-day ceiling and New Jersey's 5-year mandate. Curbing Police Misconduct and Rejecting Predictive AI (17:28) Addressing concerns regarding rogue officers abusing tracking databases—an issue highlighted by approximately 7,000 reported state database misuse cases in California in a single year—Langley acknowledged that agency manual audits reviewing only 10 to 20 searches monthly left significant blind spots. In response, Flock implemented a mandatory tool called Audit Assistance four months prior. The system analyzes search patterns to detect red flags, such as an officer repeatedly querying a specific plate without registering it on an agency-wide hot list, and directly resulted in the termination of nine police officers in Georgia for unauthorized tracking. Simultaneously, Langley rejected incorporating predictive policing algorithms that interpret vehicle movements to flag suspicious intent. He argued that public safety requires keeping humans in the loop to avoid catastrophic errors, drawing a contrast with low-stakes commercial AI applications. Flock advocates for independent third-party attestations before AI tools are integrated into law enforcement workflows. Public Backlash, Contract Cancellations, and Reversals Amid felony camera vandalism and public scrutiny on social media, Flock's subscriber churn ticked from 0% to approximately 1%, representing 60 to 70 municipal contract cancellations. Langley attributed the majority of cancellations to misinformation, citing instances where city councils conflated Flock with competitors that advertise facial recognition. However, nearly 20 cities subsequently reinstated the service after experiencing public safety declines; in Texas, suspects in an Austin shooting were arrested within 28 hours because neighboring Manor, Texas, had retained its operational Flock network. Langley and host Jason Calacanis observed that surveillance debates often reflect socioeconomic divides. While affluent homeowners frequently push back against public cameras from secure positions behind private security and gated properties, lower-income and minority populations suffer higher crime disparities and rely directly on municipal deterrence and automated investigative tools. Autonomous Drones and Municipal Transparency (49:37) Flock is shifting beyond hardware-only plate readers, which now account for just under 50% of forward-looking revenue, toward autonomous Drones as First Responders (DFR). Docked in rooftop stations at municipal stations, these drones fly autonomously at 60 mph and 400 feet, arriving on scene in under 60 seconds. Langley highlighted a case in the Dallas–Fort Worth area where a drone's 40x optical zoom revealed that an alleged firearm at a gas station was an imitation toy lighter, preventing an armed confrontation before ground units arrived. To counter fears of warrantless aerial surveillance, Flock provides municipalities with public-facing transparency portals displaying camera counts, flight patterns, and audit logs, delayed slightly to prevent active suspects from evading law enforcement. Langley noted that federal regulations govern drone airspace safety via the FAA but leave privacy rules unstandardized, leaving local governments to determine retention and oversight limits. While Flock aims to return its customer churn to zero and expand autonomous emergency response, its operating model depends heavily on municipal acceptance and self-imposed audit mechanisms in an environment marked by disparate state regulations and lingering public distrust of surveillance infrastructure.

Flock Safety 创始人兼首席执行官 Garrett Langley 在访谈中指出，车牌识别与无人机技术能显著提升公共安全治理效率，但供应商必须通过缩短数据留存期、强制实施内部审计以及明确拒绝预测性 AI，来防范权力滥用并平衡公民隐私。 规模扩张、破案实效与留存期限博弈 (5:48) Flock 成立九年已覆盖全美约 6000 个城市，系统累计协助侦破超过 100 万起刑事案件、找回逾 1 万名失踪人员。仅在 2026 年 7 月单月，系统便协助寻回 1125 名失踪人员并追踪逾 2 万辆失窃汽车；在亚特兰大，警方借助该网络在案发一小时内解救了一名后座载有两岁幼童的被盗车辆。面对扩张带来的民权争议，Langley 将数据留存定性为兼具破案资产与法律负债的双刃剑，为此 Flock 将系统默认数据留存期从 30 天大幅缩短至 7 天。 内部统计表明，约 90% 依赖 Flock 侦破的案件都在案发 7 天内解决，仅有延迟报案的谋杀或性侵等复杂案件才需更长周期。然而各州法律差异巨大：新泽西州立法强制要求数据留存 5 年，加利福尼亚州与华盛顿州设置了严格的留存上限，而新罕布什尔州则恪守传统，将车牌数据留存上限严苛限制在 3 分钟以内。为化解各地分歧并向公众解释技术局限，Flock 团队在过去一年出席了近万场市议会听证会。 技术克制与对预测性警务的明确拒绝 为消除公众对无差别大众监控的恐慌，Langley 阐述了 Flock 系统的技术边界：硬件不提供人脸识别、不进行实时流媒体直播、不拍摄车辆内部座舱，且全面禁止在车牌识别网络中针对特定个人发起查询。该系统仅索引车辆客观物理特征，例如车身颜色、制造商型号、损毁痕迹、保险杠贴纸或缺失后轮轮毂罩等线索。Langley 指出，部分竞争对手不仅公开销售人脸识别工具，还出售没有任何审计约束的脱机系统，但他明确拒绝为了商业利益牺牲伦理标准。 在人工智能议题上，Langley 坚决反对构建类似《少数派报告》式的预测性警务工具，明确表态 Flock 绝不主动通过算法去定义何为“可疑意图”。他强调，公共安全直接关乎人的生命，必须自始至终保留“人工介入”（human-in-the-loop）机制与严格的第三方客观验证。相比于商业领域 AI 生成内容带来的轻微不便，如果在 911 紧急调度中冒进引入未经审慎验证的自动化 AI，可能酿成灾难性的人命后果。 警权监督升级与强制自动化审计工具 (17:28) 面对隐私倡导组织关于“不良警察”滥用监控数据追踪私人的质疑，Langley 承认权力在任何机构中均存在被滥用的风险，例如加利福尼亚州此前一年便报告了约 7000 起州级数据库遭违规调阅的案例。传统警务监督依赖人工抽检，但大型警局日均查询量动辄达数千次，每月仅能抽查 10 至 20 次，形成了巨大的监管漏洞。为此，Flock 在数月前上线了“审计协助”（Audit Assistance）系统，并强制要求全量客户启用，不再作为可选配置。 该审计系统借助上百万起真实破案模式和历史违规滥用案例进行实时行为监测。如果一名警员频繁调阅某特定车牌，却未依规将其列入全警局共享的嫌疑车辆关注名单（hot list），系统便会触发高危警报，识别潜在的私人跟踪与骚扰行为。该功能在乔治亚州部署后，直接查实并促使相关部门开除了 9 名滥用系统的警员。Langley 主张，技术供应商不能像传统枪械或电脑制造商那样在工具售出后免除责任，而必须主动监管软件使用行为。 合约反复反转与公共安全的阶层分化 尽管社交媒体抵制以及破坏监控摄像头的联邦重罪行为频发，Flock 客户的实际解约率目前仅维持在 1% 左右（约 60 至 70 个城市）。Langley 透露，多数解约源于信息误传，例如个别城市误将竞品的人脸识别功能当成 Flock 的产品。更为显著的是，已有近 20 个曾因争议取消合作的城市重新启用了摄像头。例如，德克萨斯州奥斯汀市在停止续约后发生恶性青少年枪击案，嫌犯随后逃入临近仍使用 Flock 的马诺市，当地警方仅用 28 小时便成功缉捕三名嫌犯，直接促使奥斯汀各界重新讨论系统的必要性。 主持人 Calacanis 与 Langley 进一步剖析了关于监控权力的阶层悖论。数据显示，安全感在不同性别、种族和收入阶层中呈现极不平等的分布。居住在封闭社区、雇佣私人安保的富裕阶层往往最激烈地主张隐私权并反对技术落地；然而，在犯罪率高企、缺乏私人防护预算的工薪与少数族裔社区中，居民往往更迫切地依赖公共科技设备来遏制恶性暴力犯罪，二者对“安全特权”的认知存在深刻割裂。 第一响应无人机应用与公开透明治理 (49:37) 随着业务版图扩展，车牌识别设备在 Flock 预期营收中的占比已降至不足 50%，其增长最迅猛的业务是“第一响应无人机”（DFR）。该系统将自主无人机部署在警局或消防局屋顶的专用机箱中，接到 911 报警后能以每小时 60 英里的速度在 400 英尺高度巡航，通常在 60 秒内即可抵达现场。无人机搭载的 40 倍光学变焦摄像头能够在地面警力遭遇拥堵前传回关键情报。在达拉斯-沃斯堡大都会区的一起报警中，现场嫌疑人被怀疑持有手枪，无人机先行抵近识别确认该物体实为枪形打火机，成功避免了巡警拔枪引发致命冲突的可能。 为防范无人机沦为窥探私人领域的工具，Langley 呼吁建立更加公开透明的监管体系。Flock 当前向市政管理者提供包括飞行轨迹、机群状态和审计记录在内的“旁观者模式”（spectator mode），并向公众开放附带延时保护机制的数据门户，既确保执法行动不被泄露，又保障事后监督。Langley 指出，目前美国联邦航空管理局（FAA）仅对无人机空域安全立法，而在隐私保护领域尚无全国性标准，地方自治仍是当前平衡安全与透明度的主要治理路径。 Flock Safety 试图通过自我克制的产品架构与强制自动化审计，证明大规模公共安防技术可以在不滑向威权式面部监控的前提下运行。然而，全美在监控数据留存和无人机执法上的法律碎片化，以及公众对警察权力的长期信任赤字，表明技术界定与地方民主决策之间的博弈仍将持续。

## Background 本期内容由知名科技投资播客《All-In Podcast》于 2026 年 8 月 18 日在 YouTube 平台发布。节目由 Jason Calacanis 主持，专访了当前处于全美隐私争议风口浪尖的 Flock Safety 创始人兼 CEO Garrett Langley，深入探讨大规模监控、警察权力滥用、数据留存期博弈以及谁该掌控公共安全摄像头等核心议题。 ## TL;DR 访谈围绕公共安全与公民隐私之间的激烈博弈展开。Garrett Langley 介绍了 Flock 覆盖全美 6,000 多个城市所取得的破案与搜救成效，同时回应了因公关危机、设备破坏潮和公众对警权不信任所引发的广泛质疑。针对防止黑警滥权，Flock 推出了强制性的智能审计工具并在乔治亚州协助清退了 9 名违规警官；公司还主动将默认数据留存期从 30 天缩短至 7 天，并坚守不做人脸识别、拒绝预测警务的伦理红线。双方一致强调，监控技术的最终生杀大权必须交由民选市议会和地方社区自主决定。 ## Atmosphere & style 对话充斥着高度直率的交锋与真诚的反思，没有回避任何尖锐敏感的痛点，呈现出一种理性而紧绷的思想碰撞氛围。Jason Calacanis 以老练的投资人兼家庭父亲视角层层逼问，直戳腐败黑警、滥权监控和公关灾难的痛处；Garrett Langley 则展现出罕见的坦诚与谦逊，既痛陈沟通失误带来的内部士气挫伤，又坚定捍卫技术对弱势群体安全底线的守护价值。整场交流既有刑侦现实与法律博弈的冷峻剖析，亦有关于人权、阶层特权和科技伦理的深刻人文关怀，听感极具张力且引人深省。 ## Bulletin - [02:33](https://www.youtube.com/watch?v=5BWuvjGLMik&t=153s) Flock Safety 目前已进入全美逾 6,000 个城市，年均协助破获超 100 万起刑事案件并找回逾 10,000 名失踪人员。 - [07:15](https://www.youtube.com/watch?v=5BWuvjGLMik&t=435s) Flock 将系统默认数据留存期从 30 天大幅缩短至 7 天，这一留存窗口足以覆盖约 90% 的涉案侦查需求。 - [20:09](https://www.youtube.com/watch?v=5BWuvjGLMik&t=1209s) Flock 推出并强制启用“Audit Assistance”自动异常审计工具，已在乔治亚州协助查处并开除 9 名违规滥用系统的警察。 - [49:39](https://www.youtube.com/watch?v=5BWuvjGLMik&t=2979s) 公司第二增长曲线无人机响应系统（DFR）表现强劲，传统车牌识别（LPR）在未来预期营收中的占比已降至 50% 以下。 ## Key takeaways - [02:33](https://www.youtube.com/watch?v=5BWuvjGLMik&t=153s) Flock 从早期的社区基础车牌识别演进为涵盖车辆多维外观特征（如行李架、车贴、车身凹痕及刮擦）的综合抓拍系统，显著提升了破案效能。 - [04:22](https://www.youtube.com/watch?v=5BWuvjGLMik&t=262s) ToG 模式需要经受严密的民主问责，Flock 团队每年参与近万场市议会公开听证会，采购与部署均由地方民选官员表决裁定。 - [07:15](https://www.youtube.com/watch?v=5BWuvjGLMik&t=435s) 抓拍数据被视为“负债”而非“资产”，7 天留存期虽能解决 90% 的案件，但剩余 10% 往往涉及严重创伤或迟发性恶性犯罪，引发基层执法部门的权衡焦虑。 - [09:21](https://www.youtube.com/watch?v=5BWuvjGLMik&t=561s) 美国各州对车牌数据留存期限立法差异悬殊，从新罕布什尔州的仅 3 分钟到新泽西州的 5 年不等，Flock 主张将最终决定权交还地方社区。 - [11:00](https://www.youtube.com/watch?v=5BWuvjGLMik&t=660s) Flock 坚守“隐私优先”技术红线，坚决不做人脸识别、不录制连续视频、不拍摄车内人员，并禁止按“人”维度的搜索检索。 - [13:02](https://www.youtube.com/watch?v=5BWuvjGLMik&t=782s) 针对社交媒体煽动破坏公共安防设备的现象，破坏摄像头在多州属于刑事重罪，最高可面临 15 至 20 年监禁，且直接威胁黄金搜救窗口。 - [18:36](https://www.youtube.com/watch?v=5BWuvjGLMik&t=1116s) 公众对监控技术的抵触深层源于对警权滥用的不信任，传统人工审计日志形同虚设，促使技术向自动化异常行为建模检测演进。 - [20:09](https://www.youtube.com/watch?v=5BWuvjGLMik&t=1209s) 为杜绝公职人员私自跟踪等越权行为，Flock 上线 Audit Assistance 智能审计并设为强制启用规则，主动协助警队清除违规害群之马。 - [24:48](https://www.youtube.com/watch?v=5BWuvjGLMik&t=1488s) 科技提供商正从传统的“中立工具论”转向“主动设防”，承担起在产品底层构建强安全默认值（Strong Defaults）的责任。 - [29:30](https://www.youtube.com/watch?v=5BWuvjGLMik&t=1770s) 在 AI 警务应用上坚持“人在回路（Human in the loop）”与第三方独立审计，坚决拒绝去定义“可疑”或搞《少数派报告》式的预测警务。 - [32:42](https://www.youtube.com/watch?v=5BWuvjGLMik&t=1962s) 全美因误解退订的 1% 客户中，已有近 20 座城市在经历治安恶化或看清事实后重新启用 Flock 摄像头。 - [37:10](https://www.youtube.com/watch?v=5BWuvjGLMik&t=2230s) 宁可失去不愿接受问责监督的客户，也绝不妥协放弃系统审计与追责底线，避免因工具滥用反噬企业与社会。 - [39:15](https://www.youtube.com/watch?v=5BWuvjGLMik&t=2355s) 高净值群体享有私人安保等“安全特权”，而普通工薪阶层与弱势社区更缺乏防护壁垒，对公共科技执法与治安保障的需求更为迫切。 - [49:39](https://www.youtube.com/watch?v=5BWuvjGLMik&t=2979s) 无人机优先响应系统（DFR）在 911 接警后 1 分钟内抵达现场，利用高倍光学变焦实现精准去升级化（De-escalation），避免致命误判。 - [51:42](https://www.youtube.com/watch?v=5BWuvjGLMik&t=3102s) Flock 推出带时间延迟的“透明度门户”公开设备与航线，呼吁完善执法无人机的专项隐私法规，兼顾侦查保密与公众知情权。 - [54:49](https://www.youtube.com/watch?v=5BWuvjGLMik&t=3289s) 监控技术的最终边界应由社区自主决定，核心在于将安装、停用及留存期限的裁决权牢牢交由当地居民与民选代表掌握。 ## Detailed walkthrough ### [00:00](https://www.youtube.com/watch?v=5BWuvjGLMik&t=0s) 安全与隐私的博弈：Flock 的崛起与争议背景 主持人 Jason Calacanis 开门见山地指出，当今科技与社会交汇点上最核心的议题之一，正是“在安全与隐私之间寻求平衡”。在此背景下，站在风口浪尖的正是由创始人兼首席执行官 Garrett Langley 创立的 Flock Safety。两人曾在六七年前的播客节目《This Week in Startups》中有过深入长谈，当时 Jason 甚至曾试图投资该公司。尽管未能成为股东，Jason 多年来一直密切关注着 Flock 的演进：一方面为其业务进展所吸引，另一方面也对其可能带来的边界问题心存疑虑。 随着 Flock 在全美各大社区的大规模铺开，其在打击和遏制犯罪方面的成效有目共睹，但伴随而来的公众隐私担忧也愈演愈烈。如今的核心争议在于：“为了换取人身安全，公众究竟愿意放弃多少隐私？”当这种讨论在公众舆论与政界掀起巨大声浪时，意味着它已触及社会运作的根本神经。 --- ### [02:33](https://www.youtube.com/watch?v=5BWuvjGLMik&t=153s) 业务版图与实战成效：从社区防盗到全美六千座城市 Garrett 回顾了自己在九年前创立 Flock 的初心——仅仅是源于“生活在一个更安全的社区”这一朴素诉求。在公司创立的前两三年里，Flock 的业务完全局限于普通居民区，产品形态也是非常基础的车牌识别（ALPR）摄像头：当车辆驶过时，设备抓拍静态照片并读取车牌号。随着算法演进，如今的系统已能精准识别更丰富的车辆视觉特征，包括车顶行李架、个性化保险杠贴纸、车身凹痕以及各处损伤。 谈及实战成效，Garrett 列出了一组震撼的统计数据：仅在过去一年中，借助 Flock 系统侦破的刑事案件就突破了 100 万起；协助找回的失踪人员超过 10,000 人，其中既有患阿尔茨海默病走失的老人，也有被拐带的幼童。就在前一周的亚特兰大，一辆被盗车辆后座上坐着一名仅两岁的幼儿，亚特兰大警察局正是依靠 Flock 摄像头的实时轨迹追踪，在短短一小时内安全救回了孩子。仅在刚刚过去的 7 月份单月，系统便协助寻回了约 1,125 名失踪人员并查获了逾 20,000 辆被盗汽车。目前，Flock 的安防网络已覆盖全美 6,000 多个城市。Garrett 坦言，拥有如此巨大的社会影响力，必然伴随着公众的广泛质疑与持续审视，要在提升治安水平与捍卫个人隐私之间取得平衡绝非易事。 - **全美业务覆盖**：进入全美逾 6,000 个城镇社区。 - **年度破案与搜救**：年均协助侦破超 100 万起案件，找回超 10,000 名走失及被拐人员。 - **单月关键指标（7月）**：协助拦截 20,000 多辆失窃车辆，找回 1,125 名失踪人口。 - **技术识别维度**：由单纯的车牌 OCR 演进至车辆多模态特征识别（如行李架、车贴、碰撞刮擦痕迹等）。 --- ### [04:22](https://www.youtube.com/watch?v=5BWuvjGLMik&t=262s) 权力滥用顾虑与市政采购流程：一万场议会听证的考验 Jason 将话题引向公众最深层的忧虑——系统滥用与执法腐败风险。正如刑侦律政剧《法律与秩序》（Law & Order）或《黑道家族》（The Sopranos）中经典的影视桥段：深陷赌债的堕落警察偷偷在警务系统里违规查询车牌信息以换取利益；甚至在大型科技公司内部（如 Meta/Facebook），也曾曝出工程师利用系统特权窥探前任未加密通讯记录的丑闻。针对这类潜在滥用，Jason 追问 Flock 究竟如何从机制上防范越权，以及各地方社区究竟是如何决定引入这套监控系统的。 Garrett 解释道，面向地方政府做 ToG 业务具有极高的壁垒，这也是 Flock 早期融资步履维艰的主因——在 Flock 之前，全美已有超过二十年没有出现过专门服务于地方政府并成功上市的科技企业。每一笔公共采购合同的落地，最终的生杀大权都掌握在民选的市议会（City Council）手中。市议会通常每两周举行一次公开例会，由市政经理的下属主管（如警察局长、消防局长或公共工程部门主管）提交采购方案与概念预算。仅在去年一年，Flock 的政企团队就参加了接近 10,000 场市议会公开听证会，向公众和议员巨细靡遗地解释“Flock 究竟能做什么、绝不做什么”。 Garrett 提到，公司六年前在得克萨斯州泽西村（Jersey Village）签下的首个警局采购案，其议会答辩与表决全过程至今仍作为公开档案供全网查阅。 ```mermaid flowchart TD accTitle: Flock数据生命周期与市政决策流程 accDescr: 展示Flock摄像头从市政审批部署到数据捕获与留存策略的完整链路 subgraph Governance["民主治理与准入机制"] A["市政议会公开听证"] --> B["民选官员投票决定部署与策略"] B --> C["安装 Flock 边缘抓拍设备"] end subgraph Ingestion["特征捕获与目标比对"] C --> D["抓拍通行车辆静态高解像图片"] D --> E["提取车牌与外观特征 (行李架/贴纸/损伤)"] E --> F["实时碰撞被盗车辆与失踪通报库"] end subgraph Retention["数据留存生命周期控制"] F --> G{"留存周期策略"} G -- "7天默认策略" --> H["有效解决约 90% 常规治安案件"] G -- "到期物理销毁" --> I["最大化降低数据泄露与隐私侵权隐患"] G -- "30天及以上延长" --> J["侦破复杂凶案与迟发性重案 (占10%)"] end ``` --- ### [07:15](https://www.youtube.com/watch?v=5BWuvjGLMik&t=435s) 数据留存期之争：从三十天缩减至七天与“关键10%”案件的取舍 在数据管理哲学上，Garrett 提出了一个核心观点：“这些抓拍数据本质上是负债（Liability），而非资产（Asset）。”留存期长短直接决定了产品的破案效能：如果数据在一两天内就被彻底清除，虽然隐私风险极小，但侦查效能将大幅缩水；若留存 30 天，效能十分显著；若保留一年，效能固然更强，但隐私侵入性也会成倍攀升。Flock 过去一直将 30 天设为行业默认留存期，而近期公司做出了重大政策转向，将系统默认留存时间直接压缩到了 7 天。在科技行业中“默认设置至关重要（Defaults matter）”，但究竟保存多久、授权给谁查看，最终裁决权仍应牢牢交由地方民选官员或州法律决定。 面对 Jason 关于“数据留存时长究竟能覆盖多少案件”的追问，Garrett 给出了全美宏观数据层面的明确结论：**7 天的数据留存期足以解决大约 90% 的涉案需求**。然而，这一看似兼顾隐私与效率的取舍，却在基层执法一线引发了巨大张力。Garrett 透露，自己本周与两三百位警察局长沟通时，对方普遍感到焦虑与痛苦。因为被 7 天留存期排除在外的“那剩余 10% 案件”，恰恰是社会影响最为恶劣、媒体广泛报道的重大恶性犯罪。例如：凶杀案受害者陈尸数日后才因亲友报案请求“福利健康核查（Welfare Check）”而被发现；或者遭受严重创伤的性侵受害者，往往在巨大的羞耻感与心理煎熬中挣扎数日乃至数周，才鼓起勇气拨打 911 报警。一旦将监控线索截断在 7 天内，这类极度需要正义伸张的案件往往会因关键通行记录被自动销毁而陷入死胡同。 ### [09:21](https://www.youtube.com/watch?v=5BWuvjGLMik&t=561s) 数据留存期限与地方自主权：7天、30天与各州法律的权衡 谈及犯罪受害者（例如性侵案受害者可能需要一周甚至更久才能做好心理准备去报警）的权益，Garrett Langley 强调受害者既需要隐私，也需要施暴者被绳之以法。针对 7 天留存期能解决约 90% 犯罪这一前提，Jason 追问剩下的 10% 案件中，延长至 30 天能解决其中多少。Garrett 回应称，绝大多数剩余案件都能在 30 天内侦破。但他明确表示，Flock 不会去教各个城市该如何治理自身。现实中各州法规大相径庭：新泽西州法律规定留存期为 5 年，因此当地客户统一保存 5 年；而加利福尼亚州和华盛顿州等则立法设定了留存上限。 Garrett 认为，最核心的不是 Flock 默认设置了多少天，而是让各市议员和市长充分意识到他们拥有自主决定权。市政管理者应该与本市警察局长坐下来深入讨论，评估哪种方案最适合自身社区。哪怕某些注重隐私的社区坚持只保留 2 天或 3 天，这也远比完全没有系统要强得多，因为诸如安珀警报（Amber Alert，儿童绑架失踪警报）和银发警报（Silver Alert，老人走失警报）等分秒必争的突发危机，依然能够依靠短时数据得到侦破。Jason 对此深有同感，坦言自己夜里两点被刺耳的警报惊醒时也深感揪心却又无能为力，这种留存机制在保障紧急搜救与限制监控之间找到了务实的平衡。 - **各州法规差异**：新泽西州依法留存 5 年；加利福尼亚州和华盛顿州等均出台了法规划定留存天数上限。 - **地方裁量权**：Flock 鼓励市政官员根据当地社区偏好和治安需求，与警察局长共同敲定留存周期（即使是两三天也极具实战价值）。 --- ### [11:00](https://www.youtube.com/watch?v=5BWuvjGLMik&t=660s) 坚持不做人脸识别与车内抓拍：Flock 与行业竞品的界限 Jason 回忆起早年在节目中采访 Garrett 时的经历，当时他曾从一位父亲和注重家庭安保的角度提出疑问：难道警方不需要看清挡风玻璃后面的驾驶员长相吗？Jason 提到自己在奥斯汀的居所就部署了完备的私人监控系统（使用 Ubiquiti / UniFi 设备），其半私密领地上的摄像头带有人脸识别和车牌识别功能，部分视野甚至能拍到门前的公共道路，这在德州完全合法。许多封闭式高端社区也在出入口架设了四五个摄像头，全天候录制每一辆进出的车辆与人员。 面对行业广泛存在的全面视频监控，Garrett 划清了 Flock 与主要竞争对手的严格界限：Flock 坚决不做人脸识别、不录制连续视频、不拍摄车内人员，甚至在其车牌识别系统的底层检索逻辑中，完全禁止任何按“人”维度的搜索。许多同行正在大肆推广这些侵犯隐私的功能，而 Flock 作为行业领头羊，自然承受了最猛烈的舆论审视。Garrett 表示这并非坏事，他希望 Flock 主打的“隐私优先”（Privacy-first）架构能成为倒逼整个安防行业自我规范的标杆，让更多市议会在采购时质问竞品为何滥用技术，进而转向更克制的解决方案。 ```mermaid flowchart TD accTitle: Flock Safety与行业竞品功能边界对比 accDescr: 对比Flock与传统安防系统在人脸识别、视频录制、车内抓拍及检索方式上的差异 subgraph Flock["Flock Safety 隐私优先架构"] F1["客观车牌与车辆特征抓拍"] F2["无连续视频录制（仅抓拍事件帧）"] F3["严禁车内窥探与人脸识别"] F4["搜索权限严格限制于车牌与客观车辆线索"] end subgraph Competitors["传统安防竞品"] C1["全天候不间断视频录像"] C2["挡风玻璃透视与驾驶员抓拍"] C3["人脸比对与生物识别检索"] C4["人员跨区域轨迹追踪"] end Flock -.->|"推进行业规范"| Competitors ``` --- ### [13:02](https://www.youtube.com/watch?v=5BWuvjGLMik&t=782s) 摄像头破坏潮与重罪起诉：社交媒体煽动下的公关危机与反思 随着车牌识别系统的大规模部署，社会抵触情绪迅速升温。Jason 犀利指出 Flock 在公众沟通上遭遇了严重的滑铁卢：不仅在全美，甚至在英国也接连出现民众强行锯倒、砸毁监控摄像头的事件。反对者将此视为通向“全天候全景监控社会”的危险滑坡。Jason 坦言，自己作为父亲虽愿意让渡 30 天的车牌行驶轨迹以换取孩子的人身安全，但对人脸识别同样心存戒备。他直接批评 Garrett 此前将破坏设备者斥为“恐怖分子”的过激言论，指出这群人更准确的画像是心怀顾虑的激进抗议者（Activists），Garrett 在公关处理上显然失当，未向公众讲清 Flock 并非永久留存、不带人脸识别的技术轮廓。 Garrett 坦诚承认自己在过去一年乃至过去十年中犯下过诸多沟通错误，并直言试图列出这些教训时“清单长到让人无法向前看”。谈及现实中的设备破坏行为，Garrett 表达了深深的痛心与无奈：许多年轻人受到 TikTok 和 Instagram 上煽动性内容的挑唆，为了点赞和流量盲目效仿，却根本没意识到自己正在触犯严重的刑事重罪（Felony）。他提到俄亥俄州或伊利诺伊州最近被捕的一名涉案嫌犯，目前正面临长达 15 到 20 年的监禁。煽动情绪的网红无需承担代价，而盲从的破坏者却断送了前程；更致命的是，一旦设备被破坏，发生拐卖儿童或老人走失时的黄金搜救窗口将被彻底摧毁。 - **重罪代价**：破坏公共安防摄像设备在多州属于重罪，近期涉案个人面临最高 15 至 20 年有期徒刑。 - **社交媒体推波助澜**：线上煽动抗争与线下极端破坏脱节，抗议者误以为身处无休止的人脸大监控，实则因虚假认知以身试法。 - **民主与法治途径**：Garrett 强调，公民对监控的异议应通过市议会听证、竞选公职或正规公共辩论表达，而非实施破坏。 --- ### [16:16](https://www.youtube.com/watch?v=5BWuvjGLMik&t=976s) 安全作为基本人权：平衡隐私诉求与警权监督的挑战 在谈到部分城市因民意反弹而决定停用并拆除 Flock 摄像头时，Garrett 坦言这让人沮丧。他认为直接拆除是选择让城市变得更不安全，而明智的妥协本可以通过调整系统参数来降低采集效率，而非一刀切弃用。Garrett 指出，将隐私或安全绝对对立起来的两端都是错误的，民主程序的核心就是在复杂议题上寻求最大公约数。援引马斯洛需求层次理论（Maslow's hierarchy of needs），他认为人身安全是底层基石，也是政府对公民最根本的契约责任。 现实社会中，安全感存在着残酷的阶层与身份鸿沟： - 女性感受到的安全感显著低于男性； - 少数族裔感受到的安全感普遍低于白人群体； - 贫困社区和低收入区域的治安恶劣程度远甚于富裕街区。 Garrett 强调，Flock 的使命是弥合这种特权差距，实现安全面前的人人平等。然而 Jason 随即抛出了隐私阵营最核心的痛点：公民没有“退出机制”（Opt-out），哪怕守法出行也会被强制录入；更关键的是，如何防止体制内的腐败与警权滥用？Jason 打了个形象的比方：“如果出现一个像美剧《黑道家族》里欠了黑帮托尼·索普拉诺（Tony Soprano）十万美元赌债的黑警，他为了还债偷偷把 Flock 里的车牌追踪数据倒卖给黑帮，你们有什么审计追踪机制来阻止他？”针对这一尖锐质疑，Garrett 承认自己过去在这个问题上被批评为“过于天真”。作为一个在亚特兰大长大、自幼全心信任警察的人，他花费了整整一年时间深入接触反对者，才真正看清了公众对警察滥用权力的恐惧根源。 ### [18:36](https://www.youtube.com/watch?v=5BWuvjGLMik&t=1116s) 公众对警权的不信任与权力滥用的根源 Garrett Langley 分享了他曾遇到的一名对 Flock 抱有极度敌意女性的经历。起初 Garrett 感到困惑，强调“我们只是卖相机的”，但那位女性直接反驳道：“Garrett，我根本不信任警察，他们撒谎、欺诈、偷窃，我一个都不信。” 这让 Garrett 意识到问题的核心根源并不在于摄像头硬件本身，而在于公众对警权滥用的深层恐惧。主持人指出，这种担忧完全是有据可查的——历史上许多警局都出现过执法人员故意关闭执法记录仪、甚至执法人员戴面具且不挂警号的情况。由于公众天然对警察怀有戒心，Flock 作为警方的技术合作伙伴，自然被一同捆绑在审视的聚光灯下。 Garrett 对此表示认同，并从宏观层面剖析了权力结构的问题：无论首席执行官、民选官员还是警察，任何身处权力位置的人，滥用权力的概率都会急剧攀升。早前 Flock 系统的标准设计是提供传统的“审计日志”（Audit Log），但传统日志在实际应用中存在巨大漏洞。例如，南加州一个拥有数千名警官的大型警局，每天累计发起数千次车辆搜索，而负责审查日志的往往只有一个人。审查员每月最多只能抽查 10 到 20 条记录，面对海量数据完全无能为力，审计形同虚设。 ```mermaid flowchart TD accTitle: Flock 审计日志向自动化分析演进 accDescr: 展现从人工审查海量日志到自动化 Audit Assistance 异常检测的转变流程 subgraph Legacy["传统人工审计模式"] A["千名警官每日发起大量搜索"] --> B["海量审计日志积累"] B --> C["单人每月仅能抽查10-20条"] C --> D["绝大部分违规搜索被漏检"] end subgraph Modern["Flock Audit Assistance 智能监控"] E["实时比对正常破案与违规模型"] --> F["捕捉异常模式<br>(如连续多日单点定向追踪)"] F --> G["主动标记并推送滥用警报"] G --> H["警局领导依规问责与开除"] end ``` --- ### [20:09](https://www.youtube.com/watch?v=5BWuvjGLMik&t=1209s) 推出 Audit Assistance：揭露跟踪与滥权行为 为了解决人工审核的痛点，Flock 在四个月前推出了名为 "Audit Assistance"（审计助手）的新工具。Garrett 解释，Flock 去年协助侦破了上百万起案件，团队对正常执法的查询行为有着非常精准的基准数据；同时在过去八年里也处理过十多起滥用案例，摸清了违规搜索的特征模式。 最典型的滥用场景就是“跟踪”（Stalking）。Garrett 举例说明：如果一名警官出于个人私欲试图跟踪他人，他会在第一天输入目标车牌搜索位置，第二天再次搜索，第三天继续搜索。这种模式在正常办案逻辑中是极其反常的——因为如果是追查真正犯罪分子的车辆，警官一定会将车牌录入全系统的“布控热榜”（Hotlist），这样哪怕自己轮休或下班，其他值班警官也能收到预警；反之，若有人想暗中跟踪私事，就绝对不希望同事知情，只会进行单点隐秘搜索。 --- ### [21:20](https://www.youtube.com/watch?v=5BWuvjGLMik&t=1280s) 乔治亚州开除 9 名涉案警官与“强制启用”规则 这项新功能上线后的成效远超团队预期，直接帮助多个执法部门揪出了违纪违规的害群之马。在 Flock 的大本营乔治亚州，部分老客户警局的高层态度坚决，没有选择掩盖丑闻，而是公开通报直接开除了 9 名滥用系统进行违规查询的警察。在乔治亚州，滥用此类执法数据库属于最严重的违纪违法行为。 Garrett 对这些敢于自曝家丑、净化队伍的局长们表达了高度赞赏。他认为公众不应因此一味苛责警局，反而应该鼓励这种主动清理门户的担当。针对该工具的定位，Flock 已经做出重大调整：Audit Assistance 不再是一个可选的默认功能，而是转变为所有警局客户**必须强制启用**（Mandatory Feature）的底层规则，旨在彻底清除队伍中不配佩戴警徽的坏苹果。 --- ### [22:12](https://www.youtube.com/watch?v=5BWuvjGLMik&t=1332s) 坏苹果定律与多重审批“双钥匙”机制的取舍 主持人提出了警队中的“坏苹果概率”：哪怕假设只有 1% 的警察存在违规行为，在纽约市（约 3 万至 4 万名警官）或洛杉矶（约 1 万至 1.5 万名警官）这样的大型警队中，1% 就意味着多达 100 到 300 名害群之马，而 Flock 的审计工具在某种意义上成了捕获这些滥权者的“蜜罐”（Honeypot）。针对一线基层警官的权限控制，主持人提出是否应建立类似“双钥匙”（Double-key）或层层审批机制——即普通警员提交搜索申请及调查理由，须由值班巡佐（Sergeant）批准，再由更高级别的监督者复核。 Garrett 透露，团队的产品白板上确实已经绘制了此类多重审批机制的原型图，但推广到全美仍面临巨大的现实阻力： - **大机构 vs. 小警局**：Flock 既服务员工数千人的大都会警局，也服务得克萨斯州仅有 20 名警官、管辖 2000 人的微型小镇警局。 - **管理负担不均**：小警局警员日常极少使用该系统，审计日志非常简短，硬性推行繁重的多级审批流程会造成巨大的行政负担。 - **差异化规则**：未来系统可能针对大、中、小型城市实施分级规则，如同纽约市警察学院的培训周期远比小城镇警校更长、更严格一样。 --- ### [24:48](https://www.youtube.com/watch?v=5BWuvjGLMik&t=1488s) 技术提供商的责任演进：从“中立工具”到主动设防 全美执法数据库的滥用并非 Flock 独有，而是一个行业性的普遍问题。以加利福尼亚州为例，仅去年一年其州级数据库就报告了多达 7000 起滥用事件。《华盛顿邮报》等主流媒体对此类现象进行了深入报道。这促使 Flock 重新审视并演进了自身的责任边界。 以往科技公司的普遍逻辑是“我们只提供中立的工具，用户如何使用由他们自己负责”——类似于向警局销售笔记本电脑、监控探头、枪支或泰瑟枪（Taser）的厂商，通常不会在发生不当电击时承担直接责任。但由于 Flock 的产品直接触及公民隐私与公共安全之间的权衡，面对美国社会对个人隐私的高度关切，单纯将责任推给客户端已不再可行。 Garrett 坦言，自己对企业责任的认知已经转变：虽然最理想的情况是民选官员能够彻底通晓技术并代表社区价值观做出决策，但这并不现实（官员无法同时精通数据中心、水务系统和公共安全技术）。市长与官员依赖警察局长的专业建议，而警察局长则依赖技术提供商的引导。因此，Flock 必须主动承担起建立**强力安全默认值**（Strong Defaults）的责任，在尊重民主决策监督的同时，从技术底层筑牢防滥用底线。对话最后，主持人也引出了智能门铃品牌 Ring 创始人 Jamie Siminoff 的探索，以及消费级安防设备与警方数据协作模式的对比。 ### [27:47](https://www.youtube.com/watch?v=5BWuvjGLMik&t=1667s) 消费级安防的反思与 AI 进驻警务的边界 讨论从智能门铃品牌 Ring 曾因推出“寻狗”功能引发公众对跟踪与隐私侵犯的恐慌、随后不得不关停该功能的事件切入。Flock Safety 首席执行官 Garrett Langley 表示，他与 Ring 创始人 Jamie Siminoff 相识已久，两人并非死敌而是怀揣相同使命——让美国社区更安全。但过去一年让他们深刻体会到：科技创业者往往低估了公众对技术潜在侵害的担忧，在推进某些敏感技术时必须放慢脚步。Garrett 透露，按照过去的激进风格，Flock 本可以接二连三地快速上线各种 AI 新功能并大肆宣扬，但如今他们选择极其克制且深思熟虑地推进 AI 布局。 主持人指出，美国公众对监控技术格外敏感，容易联想到监控严密的“警察国家”模式（例如通过人脸识别无死角追踪行人、甚至监控是否乱扔垃圾或大声说话等行为）。当公众看到 Flock 与 Ring 时，难免会审视这两者与极端监控状态之间的距离。面对侦探渴望迅速破案的需求与公众隐私担忧之间的拉锯，企业必须在技术红线面前做出抉择。 ```mermaid flowchart TD accTitle: AI 与公共安全技术发展路线抉择 accDescr: 对比激进与审慎两种 AI 落地模式，展示 Flock 选择的人在回路与独立审计路径。 subgraph 激进模式["激进模式（高隐私与伦理风险）"] A1["全量上线 AI 功能"] --> B1["AI 接管关键决策 / 定义可疑"] B1 --> C1["误判增加 / 侵犯隐私 / 信任崩溃"] end subgraph 审慎模式["审慎路线（Flock 坚守原则）"] A2["审慎放缓节奏"] --> B2["坚持 Human-in-the-Loop"] B2 --> C2["第三方独立审计与客观验证"] C2 --> D2["提高破案率并降低误判率"] end ``` --- ### [29:30](https://www.youtube.com/watch?v=5BWuvjGLMik&t=1770s) 911 接警系统的 AI 风险与拒绝“少数派报告”式预测警务 在探讨 AI 的具体应用时，Garrett 举了当前争议颇大的 911 报警系统为例。民航客服（如达美航空）使用 AI Agent 即使产生幻觉，最坏的结果也不过是把下午 4 点的机票订成晚上 7 点；但若市民家中遭遇歹徒入侵并拨打 911，接听电话的 AI 却产生幻觉去询问报警人“中午吃了什么”，其后果将是致命的。部分城市出于财政预算和警力短缺的压力激进上马 AI 接警，已经酿成了严重的误判事件。Garrett 强调，Flock 并不涉足 911 呼叫业务，但明智的做法应当是只在电话线路完全占满时才让 AI 介入兜底，而非草率地“一键全盘替代人类”。 谈及利用 AI 分析汽车异常轨迹（例如低速驶向住宅又离开、反复在多家门口徘徊，判定究竟是外卖员还是踩点小偷），主持人将其类比为电影《少数派报告》（Minority Report）中预测犯罪的“预知者”（precogs）。对此，Garrett 明确划清界限：“我们绝不想成为一家去定义什么是‘可疑’（defining suspicion）的公司，那极度危险。”他强调在关乎生命的公共安全领域，必须坚持“人在回路（Human in the loop）”的原则。 - **客观性与独立验证**：AI 技术的部署必须由中立甚至持怀疑态度的第三方机构进行合规认证与测试，而非仅凭厂商或警方自说自话。 - **技术终极目标**：AI 的应用必须致力于“提高案件侦破率，同时降低抓错无辜人员的概率”，帮助警官做出更客观、更少主观偏见的判断。 --- ### [32:42](https://www.youtube.com/watch?v=5BWuvjGLMik&t=1962s) 误解导致的退订与惨痛悲剧：华盛顿州与奥斯汀枪击案的警钟 针对全美 6,000 多个合作城市中有大约 60 至 70 个城市（约 1%）选择停用 Flock 系统的现状，Garrett 坦言这十分令人痛心，而其中绝大多数退订决策都是基于不实信息的误导。他回忆道，最早选择退出的客户之一是因为误以为 Flock 具备人脸识别功能——该决策者在网上搜索“ALPR 人脸识别”时误点进了竞争对手的网站，便误以为那是 Flock 的功能而一怒之下取消了合作。Garrett 认为，向决策者和社区公众清晰传达技术的边界是公司自身的沟通责任。 更具悲剧色彩的是现实中的沉痛代价。华盛顿州某座城市在关停摄像头后的下一个周末，一名年幼的小男孩在街头遭到枪击不幸身亡，嫌疑车辆随即逃窜。由于该市关闭了摄像头，案件本可能彻底沦为悬案；万幸的是，邻近城市依然部署着 Flock 系统，最终成功抓拍到涉案车辆并逮捕了凶手。 主持人也分享了自己目前居住地德克萨斯州奥斯汀（Austin）的真实案例。奥斯汀曾在 2025 年暂停使用/未续约 Flock，随后当地发生了一起涉及青少年团伙的恶性枪击事件。嫌疑人在奥斯汀市区作案后驾车逃逸，正是邻近的马诺（Manor）警察局部署的 Flock 摄像头捕捉到了行踪，警方在 28 小时内迅速将三名年龄分别为 15 岁、16 岁和 17 岁的涉案少年抓捕归案。这一事件彻底惊醒了当地社区，推动市议会重新将引入摄像头的议题搬上台面。 - **数据留存期的各方博弈**：ACLU（美国公民自由联盟）等民权组织倾向于严格限制在 48 小时以内，警方通常希望保留 30 天，而多数社区最终设立在 7 至 14 天的平衡区间。 - **近 20 座城市重新启用**：已有近 20 个曾停用摄像头的城市在面对治安恶化与对选民的安全承诺后，选择重新开启 Flock 摄像头。 - **创业者的伦理底线**：Garrett 总结道，做企业不仅要追求规模，更要做出在多年之后能让自己的子女深感自豪的抉择。 ### [37:10](https://www.youtube.com/watch?v=5BWuvjGLMik&t=2230s) 宁失客户也要坚持追责与审计机制 Garrett Langley 明确表示，建立严格的审计与问责机制是 Flock 必须坚守的底线，即便这可能导致部分客户流失。他坦言，自己曾多次担心强制要求客户接受问责制会惹怒一部分执法部门，但他最终对团队说：“如果一个客户不希望被问责，那么我打心底里就不想要这样的客户。”如果这些机构想要没有约束、没有审计跟踪的监控系统，大可以转投竞品。Garrett 直言竞品的行事风格远比 Flock 鲁莽（reckless），没有任何约束审查机制，但他坚信“这些短视行为终究会反噬他们”。相比之下，虽然 Flock 承受了商业上的妥协与争议，但至少自己“每天晚上能睡个踏实觉”。 Jason 对此表示强烈赞同，指出不愿受约束的客户本身就是巨大的潜在公关隐患，如果由着他们滥用，迟早会引发严重的反噬效应。双方一致认为，在涉及公共隐私与执法权力的领域，缺乏边界的绝对监控工具无异于定时炸弹，宁可失去这部分订单，也不能放弃系统审计的设计准则。 ### [38:17](https://www.youtube.com/watch?v=5BWuvjGLMik&t=2297s) 超越车牌识别：以车辆外观特征破解命案 在探讨车牌自动识别（ALPR, Automated License Plate Reader）的技术细节时，Jason 询问 Flock 是否不仅抓取车牌号，还会记录车身凹痕、颜色、品牌和型号等物理特征。Garrett 确认了这一点，并解释了背后的刑侦逻辑：在突发犯罪现场，拨打 911 报警的目击者很少能在惊慌失措中准确记下完整的车牌号，人们的第一反应往往是车辆上某些显眼的特征，例如“右侧车门有撞痕”。 Garrett 分享了一个发生在亚特兰大附近的真实凶杀案：当时案发现场唯一的线索，是目击者只记得逃逸车辆的右后轮缺少了一个轮毂罩（hubcap）。办案刑警利用 Flock 系统的多维度特征筛选，检索了案发前后 15 分钟内经过该区域的所有缺少右后轮轮毂罩的车辆。这一线索迅速锁定了嫌疑目标，为侦破整个凶杀案打开了突破口。Garrett 特别强调，这种基于特征的搜索“只是侦查的起点，而不是调查的终点”，但正是为一线优秀刑警提供了最初的关键拼图，才得以将凶手绳之以法。 ```mermaid flowchart TD accTitle: Flock 车辆多维特征侦查流程图 accDescr: 展示从目击线索到 Flock 特征筛选，再到落地侦查的刑侦闭环。 subgraph S1["案发现场线索获取"] A["突发刑事案件发生"] --> B["911 报警与目击者证词"] B --> C["模糊/非车牌特征<br/>（如右后轮缺少轮毂罩、特定凹痕、颜色）"] end subgraph S2["Flock 平台多维筛选"] C --> D["设定案发时间窗口<br/>（例如 15 分钟内）"] D --> E["特征匹配算法检索<br/>车型 / 破损 / 颜色 / 缺失部件"] E --> F["输出高相关候选车辆列表与图像"] end subgraph S3["刑侦落地与破案"] F --> G["侦查员核查背景与行动轨迹"] G --> H["成功锁定犯罪嫌疑人"] end ``` ### [39:15](https://www.youtube.com/watch?v=5BWuvjGLMik&t=2355s) 旧金山犯罪率逆转背后的“Flock 效应”与特权视角 谈到旧金山近年来的治安变化，Garrett 提到了社交媒体上讨论的“Flock 效应”（Flock Effect）——旧金山的暴力犯罪和机动车盗窃率均出现显著下降。Garrett 谦逊地表示，Flock 只是这一巨大转变中的一小部分，真正的核心在于更强硬的民选官员、更严格的执法措施、强有力的警察局长、态度坚决的地检官（DA），以及整个社区对安全环境的迫切要求。“如果你住在旧金山，你理应享有安全，它是全美最好的城市之一。” 随后，讨论转向了对“治安与隐私”之争的社会学剖析。旧金山长期深受嬉皮文化、“爱之夏”（Summer of Love）与自由意志主义熏陶，崇尚“顺其自然、特立独行”（live and let live / let your freak flag fly）。这种文化甚至一度演变成对盗窃犯罪的过度宽容和心理学借口。然而，Jason 尖锐地指出，很多站在道德高地反感治安监控的人，往往忽略了自身的“安全特权”（security privilege）： - **富人阶层的防护壁垒**：居住在太平洋高地（Pacific Heights）等富人区的高净值人群，家中配有全套私人监控系统、门卫、保镖乃至私人司机；一旦街区出现形迹可疑的流浪汉或陌生车辆，往往会瞬间引来数十名警察与私人安保围堵。他们天然享有安全，因而可以轻描淡写地高谈隐私。 - **底层社区的真实困境**：普通工薪家庭无法承担封闭式社区（gated community）的昂贵开销，门铃监控和公共摄像系统往往是他们免受侵害的唯一防线。真正长期承受犯罪侵害的群体，反而是最渴望安全与科技执法的一方。 ### [41:03](https://www.youtube.com/watch?v=5BWuvjGLMik&t=2463s) 拦阻搜查历史反思与社区安全的权衡取舍 Jason 结合自己家庭的经历，回顾了纽约市当年的“拦阻搜查”（Stop, Question/Ask, and Frisk）政策。他的兄弟曾是该时期的纽约警察。尽管该政策后来因侵犯宪法第四修正案而引发巨大司法争议并被叫停，但在 20 世纪 70 至 80 年代暴力与枪支泛滥的纽约街头，治安环境极为恶劣。Jason 指出，媒体报道往往刻意省略了中间的“Ask（盘问）”，实际上警察常常是通过主动询问来摸排危险人员；而最具讽刺意味的现实是，当年最支持拦阻搜查的，恰恰是治安最差、饱受暴力折磨的底层贫困社区。 Jason 认为，在面对监控技术时，人们的态度高度取决于个人所处的人生阶段与阶层： - 没有孩子、独自居住的年轻群体，往往将个人隐私置于公共安全之上； - 拥有住房、承担养育责任的父母群体，则几乎无一例外地将家庭与社区的安全放在首位。 对于技术边界，Jason 明确划出了红线：“我不希望政府建立一个全国性的 Flock 数据库，我也坚决反对政府推行全方位的面部识别（facial recognition），我认为那跨越了红线。”但他同时坦承，作为一名父亲和社区成员，他完全接受地方社区保留 14 天到 30 天的车牌数据，“如果有人因此对我感到失望，我很抱歉，但我能接受车牌识别系统存在 30 天，这是我深思熟虑后的结论。” ### [44:39](https://www.youtube.com/watch?v=5BWuvjGLMik&t=2679s) 伊利诺伊州反对者的转折与公众舆论的修偏 针对技术接受度的转变，Garrett 讲述了一桩发生在他与伊利诺伊州一位民选官员之间的真实故事。该官员曾是 Flock 最激进的公开反对者与抨击者之一。然而极具讽刺且令人后怕的是，她随后遭遇了政治暴力，住所遭到枪击。危急关头，正是当地警方利用安装在附近的 Flock 摄像头，迅速追踪并抓获了开枪袭击的嫌疑人。 经历了这场惊魂未定的袭击后，这位官员公开改变了立场，从昔日最激烈的抗议者变成了 Flock 最坚定的支持者之一。Garrett 感慨道：“你无法选择自己何时成为暴力的受害者。有时暴力是随机的，有时是蓄意的，但当灾难降临时，每个人都渴望正义，每个人都想确信自己生活在安全之中。”他强调，公共讨论应当建立在务实平衡的基础上，绝不能让极少数声音最大、享有安全特权的群体，主导整个社会走向更不安全的境地。 ### [45:35](https://www.youtube.com/watch?v=5BWuvjGLMik&t=2735s) 面对公关风暴与公司内部士气的真实挑战 Jason 随后切入了一个极具火药味的话题：过去一年里，Flock 陷入了外界普遍认为的“公关灾难”（PR disaster），遭遇了全方位的舆论围剿与隐私质疑。他直接追问 Garrett，这场公关危机是否足以致命，以及公司内部受到的真实创伤究竟有多深。 Garrett 坦承，持续不断的公关风波带来的最严重创伤其实是“公司内部的士气”（internal morale）。Flock 的团队成员在 X（Twitter）或 Reddit 上看到排山倒海的愤怒与谩骂时，内心往往感到困惑与委屈：“员工们会忍不住去想，我们过去九年里一直在做完全相同的事情、践行着相同的安全使命，为什么突然之间有这么多人对我们恨之入骨？”这种外部舆论与公司内部使命感之间的撕裂，成为了过去一年管理层必须直面的最大心理考验。 ### [46:22](https://www.youtube.com/watch?v=5BWuvjGLMik&t=2782s) 面对外界质疑、客户纽带与 1% 流失率背后的安全核心诉求 面对外界强烈的舆论压力与误解，Garrett Langley 坦言，当团队明明一心只想保护社区安全，却因为一些自己根本没做过的事情而遭受铺天盖地的攻击时，这种从心底涌出的挫败感是非常沉重的。但从另一个角度来看，这场危机反而拉近了 Flock 与执法客户之间的距离。对于前线警员和警长而言，维护社区安全是他们的职责，甚至有警长直言 Flock 的技术是“继 DNA 技术以来打击犯罪最伟大的发明”。商业层面上，Flock 年初完成了新一轮融资，业务仍在持续增长；尽管客户流失率（churn）从过去的 0% 微升到了 1%，但 Garrett 表示他的目标是全力争取将其重新降回 0%，方法就是更主动地沟通、更认真地倾听。 主持人将安全需求与马斯洛需求层次理论（Maslow's hierarchy of needs）相提并论，指出安全感与食物、居所一样是人类无法妥协的最底层需求。因此，科技企业的领导者必须走在舆论与技术风险的前面，而不是一味被动防守。主持人以扎克伯格（Mark Zuckerberg）及 Meta 过去在 Instagram 青少年心理健康、社交成瘾问题上的被动应对为例，认为如今各家科技 CEO 在面对 AI 与监控技术的伦理争议时，必须展现同理心并主动制定规则，通过开源与普惠工具来拉平竞争环境，而不是等到公众恐慌发酵后再四处灭火。 - **业务基本面**：公司年初完成融资，整体处于增长态势；虽然客户流失率从 0% 变为 1%，但管理层正通过增强沟通与透明度力争重回零流失。 - **用户黏性与价值认同**：基层警局将该系统视为不可或缺的执法利器，称其对破案的帮助堪比现代法医 DNA 技术。 - **舆论与媒体应对**：面对外界诸如 Tucker Carlson 等保守派媒体人的猛烈抨击，Garrett 表示 Flock 积极登上《今日秀》（Today Show）与《早安美国》（Good Morning America）等大众媒体，直接向公众普及事实。 --- ### [49:39](https://www.youtube.com/watch?v=5BWuvjGLMik&t=2979s) 拓展第二增长曲线：无人机响应系统与德州加油站惊魂案例 无人机（Drones）已成为 Flock 增长速度最快的业务板块，公司业务结构正在多元化，传统的车牌识别（LPR）业务在未来预期收入中的占比已降至 50% 以下，其余则由无人机、实时监控摄像头及智能软件构成。Flock 的无人机日常部署在警局或消防局屋顶带有温控（HVAC）的专属停机坞（Dock）中。一旦接到 911 报警电话，系统可以在派遣地面警车之前率先出动无人机（Drone as First Responder, DFR）。 ```mermaid flowchart TD accTitle: 911 无人机优先响应与化解危机流程 accDescr: 展现 911 报警后无人机如何先于地面警力抵达现场并精准识别风险以避免暴力冲突 subgraph Dispatch["报警与部署阶段"] A["民众拨打 911 报警<br/>（报告潜在持枪威胁）"] --> B["屋顶温控停机坞（Dock）<br/>自动弹射无人机"] end subgraph Aerial["空中巡航与侦察"] B --> C["400 英尺高空巡航<br/>时速 60 英里直线飞行"] C --> D["40x 高倍光学变焦镜头<br/>超视距高清侦测"] end subgraph Resolution["现场判定与危机化解"] D --> E{"目标物核实"} E -->|"非真实枪支"| F["确认为玩具打火机"] F --> G["调度员电话安抚报警人<br/>通知地面警员解除致命武力警戒"] end ``` Garrett 分享了一个发生在达拉斯-沃斯堡（Dallas-Fort Worth）地区的真实案例：深夜一名女子在加油站报警称一名男子在周围徘徊并“挥舞枪支”。在得克萨斯州，公开携带枪支虽不违法，但挥舞枪械却会引发恐慌。按照传统的执法流程，警员在高度紧张状态下拉响警报狂飙到现场，极易演变成拔枪射击的致命悲剧；但这一次，警方先行派出了无人机。无人机在 400 英尺高空以每小时 60 英里的速度沿直线飞行，不到 1 分钟便抵达现场，并利用 40 倍光学变焦镜头拉近观察。画面显示，该男子手里拿的根本不是枪，而是一个外形逼真的玩具打火机。调度员随即回电安抚报警女子解除警报，成功化解了一场一触即发的流血危机。Garrett 强调，安全不仅是抓捕罪犯，更在于利用技术去化解潜在冲突，让整个社区获得实质的安全感。 - **多元化业务架构**：LPR 业务预期营收占比已降至略低于 50%，无人机、实时视频与软件成为核心增长极。 - **DFR 技术指标**：无人机驻扎于屋顶温控坞，400 英尺巡航高度，时速 60 英里（约 96 公里/小时），搭载 40 倍光学变焦镜头，可在 1 分钟内赶赴现场。 - **去升级化（De-escalation）效果**：通过超前获取高清现场画面，避免地面警员在信息模糊与高度应激下误用致命武力。 --- ### [51:42](https://www.youtube.com/watch?v=5BWuvjGLMik&t=3102s) 飞行监管、审计追踪与透明度门户的制度构建 针对无人机可能引发的隐私侵犯担忧（如偷窥窗户或被不良警员用于跟踪私人关系），主持人探讨了引入类似安永（Ernst & Young）等第三方审计机制的必要性。Garrett 指出，执法机构在引入无人机时必须吸取当年随身执法记录仪（Body Cameras）普及过程中的教训，建立更为严密的监督体系。目前在地方政府架构中，城市经理（City Manager）等文职官员承担着对警察局长的监督职责，他们可以直接查阅系统的所有审计日志与飞行记录。 目前 Flock 为绝大多数客户配备了面向公众的“透明度门户”（Transparency Portal），公开展示摄像头数量、无人机设备以及历史飞行路线等数据。为了防止泄露正在进行的犯罪侦查行动，飞行轨迹的公开会设置一定的安全时间延迟。此外，系统还设有专供市政官员和指挥层实时监督的“观战模式”（Spectator Mode）。Garrett 指出，尽管公众对透明度和问责制的要求不断提升，但目前全美除联邦航空管理局（FAA）的基础空域安全规定外，在州和城市层面对执法无人机的隐私规范仍几乎处于法律空白状态；这种立法滞后与车牌数据保留期的巨大差异如出一辙——在激进的新罕布什尔州，车牌数据只能保留 3 分钟，而在纽约州则允许保留 21 天。 - **监督机制与角色**：城市经理拥有调阅审计日志与无人机飞行轨迹的独立权限，对警局权力形成制衡。 - **透明度门户设置**：公开摄像头清单、无人机配置及带时间延迟的历史飞行路线，兼顾公众知情权与前线刑侦保密需求。 - **法规空白与地域差异**： - **无人机监管**：缺乏除 FAA 基础空域安全之外的州级或联邦隐私专项法规。 - **车牌数据保留期对照**：新罕布什尔州法律极为苛刻，仅允许保留 3 分钟数据；纽约州则允许保留长达 21 天。 ### [54:49](https://www.youtube.com/watch?v=5BWuvjGLMik&t=3289s) 社区自主权与离网隐私生活：访谈总结 在节目的尾声，主持人借着前面关于各州法规的讨论打趣道，如果听众真的想要彻底与世隔绝、过上完全避开监控的“离网”（off-the-grid）隐私生活，也许可以考虑搬去华盛顿州，或是崇尚“不自由毋宁死”（Live Free or Die）的新罕布什尔州。随后，主持人向加勒特·兰利（Garrett Langley）表达了诚挚的谢意，赞赏他面对尖锐问题时的坦诚直率，并直言在如此充满争议与隐私敏感度的领域创业并试图主动解决行业痛点，确实需要巨大的勇气与定力。 主持人总结指出，公众完全可以在具体细节和边缘问题上持有不同看法与争论，但整场讨论中最核心的关键点（"the high order bit"）在于：这项技术的最终控制权必须掌握在社区自己手中。Flock 当前正在走上正确的轨道，其核心理念正是把决定权交还给当地居民——如果一个社区不希望被摄像头覆盖，他们完全可以选择停用并关闭设备；如果他们认为数据存储时间应当压缩到 48 小时，系统便可以严格限定在 48 小时，并在规则内设立合理的上限。随着主持人送上祝福并感谢听众的收听，本期《All-In》专访在标志性的片尾旋律中正式画上句号。 - **核心结论（The High Order Bit）**：公共安全与监控技术的边界应由社区自主决定，包括是否安装、何时停用，以及数据保留时长（例如缩短至 48 小时）等具体策略。 - **隐私逃避选项的调侃**：主持人笑谈如果追求绝对的离网隐私，新罕布什尔州和华盛顿州是代表性的避风港。 ## Video description (0:00) The most controversial company in privacy right now, Flock CEO joins the show! (7:23) License plate data retention: 7 days solves 90% of crimes (13:00) Camera vandalism, felony charges, and privacy concerns (18:15) Dirty cops exposed: Flock's audit tool got 9 Georgia officers fired (28:25) AI, drones, facial recognition, and avoiding predictive policing (39:43) Safety is a privilege and who actually needs Flock (45:36) Flock’s PR crisis: internal morale, churn, and 20 cities turning the cameras back on Thanks to our partners for making this possible! AppLovin Ads is AppLovin's AI advertising platform reaching over a billion daily active users across mobile games. Full-screen video ads with a 35-second median watch time. Advertisers are profitably spending hundreds of thousands of dollars a day. AppLovin Ads is now open to all advertisers. Sign up at http://applovin.com/ALLIN Numeral is the trusted solution for U.S. sales tax, VAT, and GST, used by 3,000+ businesses globally. They handle registrations, filings, and tax rates, so you can stay ahead of risk. Learn more at https://www.numeral.com/?utm_source=all-in Follow Garrett: https://x.com/glangley Follow the besties: https://x.com/chamath https://x.com/Jason https://x.com/DavidSacks https://x.com/friedberg Follow on X: https://x.com/theallinpod Follow on Instagram: https://www.instagram.com/theallinpod Follow on TikTok: https://www.tiktok.com/@allin Follow on LinkedIn: https://www.linkedin.com/company/allinpod Intro Music Credit: https://rb.gy/tppkzl https://x.com/yung_spielburg Intro Video Credit: https://x.com/TheZachEffect #allin #tech #news [Source](https://www.youtube.com/watch?v=5BWuvjGLMik)

[5:48] Garrett Langley: This data is a liability to an asset. You should have a set retention period, and that retention period for how long this data will be stored is going to determine the efficacy of the product. [38:09] Garrett Langley: I remember so vividly we had this homicide case near me in Atlanta and all the person could remember was that the back right tire was missing a hubcap... Detectives went in, filtered the 15-minute window, and found the person.

[0:00] All right everybody, welcome back to the [0:01] allin interview. This is where we find [0:04] the people building the future. [0:05] Sometimes a politician, sometimes a CEO, [0:08] sometimes an investor. All the time an [0:10] important discussion and one of the most [0:12] important discussions of the moment is [0:15] balancing safety and privacy. And there [0:18] is one company that has, you know, is at [0:21] the center of this. And we're so lucky [0:23] to have Garrett from Flock here, Garrett [0:25] Langley. Garrett, you and I spoke on [0:27] This Week in Startups, my other podcast [0:29] that I've been doing for 16 years. I [0:30] don't know how many years ago that was, [0:32] six or seven years ago. [0:33] >> Six or seven years. Yeah. [0:34] >> And I've been following your company. [0:36] Uh, full disclosure, I was like trying [0:37] to invest at one point. I I'm not a [0:39] shareholder, but I've always been [0:41] intrigued by what you're doing on the [0:44] margin sometimes concerned. And it seems [0:47] like you've been so successful with [0:49] Block, which does safety cameras, [0:52] license plate readers specifically. [0:53] We're going to get into all the details [0:54] today, but in the last year because of [0:57] the tremendous success of Flock, it has [1:00] been adopted in many communities. We're [1:03] going to get into that. And it's had a [1:04] really great impact at reducing crime. [1:08] We'll get into that. But there is now a [1:11] group of people who maybe are have [1:13] privacy concerns. And you and I have [1:15] talked over the years, and we're going [1:17] to get into it. We're going to go right [1:18] at the controversies and everything and [1:21] this grand debate we're going to have. [1:22] How much privacy am I going to give up [1:24] to get safety? And and it's really [1:26] interesting to me that there's been so [1:29] much energy on this recently because [1:31] when I see that kind of energy as an [1:33] investor um or as a podcast host here on [1:36] the world's greatest moderator TM, I [1:38] know there's something at stake. [1:43] >> App started with an $8 domain and no VC [1:46] funding and became one of the largest ad [1:48] platforms in the world. Now, that same [1:50] engine powers app 11 ads for e-commerce. [1:53] Your ads run inside mobile games, [1:55] reaching over a billion people with full [1:57] screen distraction-free attention. The [1:59] platform finds buyers and optimizes for [2:01] profit. You set the target, it does the [2:04] rest. One cookware brand went from 4 [2:06] million to $16 million, turned [2:08] profitable, and is on pace for 80 [2:10] million this year. Visit [2:11] apploven.com/allin [2:13] to launch your first campaign today. I'm [2:16] going all in. [2:18] >> Tell us a little bit about flock safety [2:20] and how many communities you're in now [2:24] and then also the drone component [2:26] because people know the license plate [2:28] readers, maybe they don't know the [2:29] drones. So, let's just get that [2:30] footprint and build a mental model for [2:32] the audience here. [2:33] >> That's great. Yeah, Jason, thanks for [2:35] having me. Um, so I started the company [2:36] nine years ago with a pretty simple [2:38] mission, which is to live in a safer [2:40] neighborhood. And that was our initial [2:42] pitch. You know, if you look at our [2:44] first two or three years in business, it [2:46] was just neighborhoods. We sold a [2:47] neighborhood, so we sold a license plate [2:49] reading camera, which is pretty basic [2:52] technology. A car drives by. We take a [2:54] still image of vehicle. We read the [2:56] license plate. And the system's gotten [2:58] more sophisticated with time. We can [3:00] detect things like a roof rack or a [3:02] bumper sticker or dent or damage, but [3:05] that's at at its core what the [3:06] technology does. And if you look at like [3:08] last year, to your point, we've had a [3:10] tremendous impact on the country. um [3:12] north of a million crimes solved using [3:15] Flock. North of 10,000 people found who [3:18] were missing. And that's everything from [3:19] our grandparents with dementia to our [3:22] children. Like just last week in [3:23] Atlanta, there was a stolen car that [3:25] happened to have a 2-year-old in the [3:27] back seat. And thankfully, because of [3:30] very hard work by Atlanta Police [3:32] Department and our cameras, that little [3:33] kid was safely returned within the hour. [3:36] I don't know what that would have [3:37] happened if without Flock. A and that's [3:40] one story, but that's my own city last [3:42] week. And you look at July as a, you [3:45] know, as a data point, last month it was [3:47] about 1,125 [3:50] missing people and over 20,000 stolen [3:52] cars. And those are the cars we use to [3:54] drive our kids to school, to drive to [3:56] work. And so we're in, you know, roughly [3:58] six, you know, just over 6,000 or so [4:01] cities um across the country. And it and [4:04] that the consequence, you know, of being [4:06] that consequential is people have a lot [4:08] of questions and people have a lot of [4:09] debates. And I think it's a it's a [4:11] worthwhile thing to talk about what kind [4:13] of society do we want to live in? I want [4:15] to live in one where we're safer. Um, [4:17] and I want to keep my privacy while I do [4:19] it. But that is hard. You know, to find [4:20] that balance is not easy. [4:22] >> I remember when we had our first [4:24] discussion, gosh, so many years ago on [4:25] this week in startups, we were talking [4:28] about it and I had brought up this, you [4:30] know, hey, well, what happens about [4:31] abuse in the system? Because once in a [4:32] while you get a dirty cop. If you watch [4:34] any kind of procedural TV, you watch Law [4:36] and Order, you watch the Sopranos, the [4:38] Sopran, you know, Tony Sopranos got a [4:40] dirty cop who's, [4:41] >> you know, you know, in the hole for, you [4:44] know, some gambling debt and they run [4:46] license plates for them, right? This is [4:47] like a trope on the news. And it's a [4:50] trope because it's common, [4:53] >> not like on a percentage basis, but we [4:55] all know that any system that records [4:58] data could be abused. We saw it at [5:00] Facebook. You had engineers there [5:02] famously stalking their ex [5:06] >> and like reading their unencrypted [5:09] messages and the industry had to make a [5:11] lot of changes. So I remember at the [5:13] time you were telling me, hey, the [5:15] settings and this is where we're going [5:16] to get really granular today. The [5:18] settings are set by the local community [5:20] and these cameras are installed by a [5:22] local community. The communities I've [5:24] lived in have had them. So let's answer [5:27] these two questions. Number one, how [5:30] does a community wind up having flock [5:33] cameras installed? What's that decision [5:36] making process like typically? And then [5:38] we'll get into very granularly what [5:40] you're recording, for how long, who has [5:43] access, the audit logs, and then what's [5:47] not recorded. [5:48] >> Yep. No, that's a it's a great question. [5:50] So, you know, one one of the reasons uh [5:53] Jason we struggled so much to raise [5:55] money as a business is because it is [5:58] traditionally and was for flock very [6:00] hard to get contracts done with local [6:02] government. You know, the the last [6:04] company to go public that serviced local [6:06] government was over 20 years ago. Um, [6:10] this is a really hard space because city [6:12] councils have the final decision [6:14] >> and so every two weeks a city council [6:17] meets and they're, you know, the city [6:19] managers, direct reports, be it a police [6:21] chief, a fire chief, uh, public works, [6:24] presents ideas, presents, uh, concepts [6:27] they would like to get approval for. And [6:29] Flock is on that list. And so I think [6:31] last year we went, we went to something [6:33] close to 10,000 city council meetings as [6:35] a business. I have a huge team whose [6:38] entire job is just going to city council [6:39] meetings explaining what flock does and [6:41] what flock doesn't do. Um and and it's [6:44] just a it's just the cost of doing [6:45] business with local government. Um so [6:47] it's really straightforward. Um and I [6:49] was actually talking to the chief our [6:50] first ever police chief down in Jersey [6:52] Village, Texas. I was talking to him a [6:53] couple weeks ago and like that video [6:56] that council vote six years ago is still [6:59] public record. [7:00] >> Yes. City council meetings are famously [7:03] recorded. [7:04] >> It's tough. It's a wacky environment. [7:06] >> It it is it is it can be particularly [7:09] wacky and but we we go right because we [7:11] want people to know what they're buying [7:12] and so they they buy the product and you [7:15] know what has been true for the last [7:17] nine years um is we we've always said [7:20] this is this data is a liability to an [7:22] asset. You should have a set retention [7:24] period and that retention period for how [7:27] long this data will be stored is going [7:29] to determine the efficacy of the [7:30] product. And so if you delete this data [7:33] in a day or two, it's going to be less [7:35] effective. It's still going to be better [7:36] than nothing. And if you hold this for [7:38] 30 days, it's going to be pretty [7:39] effective. And you hold it for a year, [7:40] it'd be more effective. And we always [7:42] believed city councils should have that [7:44] final say for how long they want to [7:46] store it and who they want to share it [7:48] with. And so for the longest time, 30 [7:51] days was our default. We've recently [7:52] shifted that down to 7 days. Um we think [7:55] that's a reasonable trade-off, but we [7:57] still think cities should should decide [7:59] um or states for that matter. But [8:01] someone who is democratically elected [8:03] should have the final decision of how [8:05] this technology works in their [8:06] community. But what I think you probably [8:08] agree with as a technology provider, we [8:10] have to set strong defaults. We default [8:12] matter is what we say in the industry. [8:15] Yes. And you know if you think about the [8:17] number of crimes of a hundred crimes, [8:20] how many would and I'm sure you have [8:22] some idea of this. If we have a hundred [8:25] crimes that occurred of whatever the [8:26] nature are, how many are solved with [8:29] seven days of data, 30 days of data or [8:33] require more than 30 days of do you have [8:35] that? [8:36] >> We have a we have a good sense, right? I [8:38] mean, I'm going to give you an aggregate [8:39] data point across the whole country. [8:42] >> We think 90% of crimes will get solved [8:44] within seven days of data retention. [8:46] >> Beautiful. [8:47] >> It is, Jason. But I would tell you I I I [8:49] probably talked to two or 300 police [8:51] chiefs this week and and they remind me [8:53] and and I tell them you should tell your [8:54] city council this too. [8:55] >> Yeah. [8:56] >> The 10% we're talking about or the 10% [8:58] you read about in the news though. [9:00] >> Oh, right. It's the it's the homicide [9:03] that there was never a 911 call because [9:05] someone was killed and it takes five [9:08] days to discover the body because of a [9:09] welfare check and then it's it's a [9:11] complicated case or it's look and I'm [9:13] sorry to you know go to this topic. It's [9:16] a rape victim who felt guilt, who felt [9:18] shame, and she didn't call 911 until a [9:21] week later because she wasn't ready. [9:24] >> Yes. [9:24] >> And that's real life, right? Like I I [9:25] deal with this every day, so I'm [9:27] probably less sensitive to it. But these [9:29] victims deserve privacy. They deserve [9:32] accountability for the people who attack [9:34] them [9:35] >> to even identify the crime. So what [9:37] you're saying is now would 30 days of [9:41] the 10% remaining that you know we're [9:43] just ballparking here [9:44] >> but that last 10% how many of those [9:47] >> out of 10 would be resolved with 30 days [9:50] of data? We think the majority and look [9:52] we're not going to lose we're not gonna [9:54] we're not going to tell cities how to [9:56] run themselves and to be clear you know [9:59] there's states like New Jersey where [10:00] state law is 5 years and so every [10:02] customer holds it for 5 years in the [10:04] state of California there's a max [10:06] retention in the state of Washington you [10:08] know most states have now adopted [10:10] regulation [10:12] >> and they set a maximum limit but I think [10:14] for me Jason what's important more than [10:17] my just default is making sure city [10:20] councilor s and mayors are educated that [10:22] they have control. They can make this [10:25] choice and they should actually go spend [10:26] time with their police chief and figure [10:28] out what does he or she think is the [10:30] best for that community. And I think [10:32] look, there's some communities who are [10:34] going to push for two days or three [10:36] days. And I think that's that's so much [10:37] better than nothing because that Amber [10:39] Alert and that silver alert, which are [10:40] the things we all lose sleep over, y [10:42] >> they're still going to get solved, too. [10:45] >> I literally lose sleep over. I got one [10:46] at 2 in the morning the other day and [10:49] you know I wake up, I read it. I'm like [10:50] I'm not I don't know what I'm going to [10:52] do with this data at 2 a.m. I'm in bed. [10:54] But okay, so this seems completely fair. [10:57] Um in terms of how you're expressing it. [10:59] >> Um and to be clear, when I had you on [11:02] Twist back in the day, I was like, well, [11:04] don't you need to know who's in the [11:06] driver's seat and wouldn't you want to [11:07] get a picture of the windshield? And I [11:10] was kind of finding myself advocating in [11:12] my community, hey, I would be okay just [11:15] as Jay Cal the dad and somebody who [11:18] sometimes on the margins has some, you [11:20] know, privacy security issues. They're [11:22] not acute like for for, you know, I [11:24] don't know, Sam Alman tragically and and [11:27] what's happened at his house twice, you [11:29] know, but [11:29] >> I have my own security cameras. I use uh [11:33] Unity. It's incredible. [11:34] >> Great product. [11:35] >> And I have a license plate reader on [11:37] mine. I have facial recognition on my [11:40] semi-private property. Of course, some [11:42] of it does look out on public roads [11:44] because I have access roads. So, I am [11:47] technically able to record the license [11:49] plate of everybody who drives by the [11:50] front of my house. I think that's legal [11:52] to do here in Austin. [11:53] >> Yes, it is legal. [11:54] >> So, are there communities that are using [11:59] other systems that are not flock and [12:02] recording every single person in and [12:05] out? Because when I go in and out of [12:06] private communities once in a while, I [12:08] see there's four or five cameras. I know [12:09] I'm on camera. So, aren't people already [12:13] doing full video in their communities, [12:16] not using? [12:17] >> Yeah. No, Jason, you're spot on. I mean, [12:20] we don't do facial recognition. We don't [12:21] capture video. Um, we don't look inside [12:24] the car. Um, we don't even allow in our [12:27] system to search for people um for our [12:30] license cameras. And that's in direct [12:32] conflict with some of our biggest [12:34] competitors who do all of the things I [12:36] just mentioned. [12:37] >> And unfortunately um or fortunately, [12:40] we've been very successful. And as kind [12:43] of the industry leader, the spotlight's [12:45] on us. And and that's okay because my [12:47] hope is the changes we're making, the [12:49] changes we'll continue to make will [12:50] force the rest of the industry to follow [12:52] suit. And I hope counselors ask [12:54] questions of the systems they have if [12:55] it's not flock. And maybe they'll switch [12:57] to Flock if they want a privacy first [12:59] solution. And you guys have had a bit of [13:03] a communication challenge. People were [13:06] ripping down the cameras. There's an, [13:08] you know, people care about privacy. Not [13:10] just here in the United States, by the [13:11] way. In the UK, people are ripping down [13:13] cameras. There are a group of people who [13:15] are entitled to this opinion. And I [13:17] share it on the margins. When I had [13:20] children, it changed dramatically [13:22] because number one for me is security. [13:25] I'm willing to give up my license plate [13:26] data. If somebody knows my exact [13:28] whereabouts for 30 days of my license [13:30] plates, I'm okay with that [13:32] >> because I have children and I'm okay [13:35] with giving up that amount to privacy. I [13:37] would be a little concerned, I think, [13:38] with the um with facial recognition if [13:41] I'm being totally honest. [13:42] >> But, you know, it's kind of a moving [13:44] target. And I guess the reaction you [13:48] guys had was, and I think I I had talked [13:50] to you about this, is you called them [13:51] terrorists, like the people taking the [13:53] cameras down and you know, this kind of [13:55] thing. Where does this audience who's [13:57] ripping down cameras? I mean, they're [13:59] obviously not terrorists. Um, I think [14:01] they're like activists, I would say, [14:04] would be a better descriptor. Um, but [14:07] you guys haven't communicated this [14:09] perfectly in my opinion. Correct. So, [14:11] where are people coming to you with [14:14] valid concerns? Yeah. [14:15] >> And then how have you evolved your [14:17] ability to communicate what you're [14:19] doing? And that's one of the reasons I [14:20] wanted to have you on this is because I [14:21] looked at and I was like, I don't think [14:23] you're doing a particularly good job of [14:24] communicating this. And I also think the [14:26] people who are ripping down the cameras, [14:29] they think they're in a surveillance [14:30] state with permanent records and facial [14:33] recognition. So I think they don't [14:34] actually understand that the nuances and [14:36] contours of the product. So talk about [14:38] where you got it right, where you got it [14:39] wrong, and how you've evolved in [14:41] communication. [14:42] >> Yeah. No. And look, I'm far from [14:44] perfect. And I've made my fair fair [14:46] share of mistakes over the last year and [14:48] good god, over the last decade, I [14:50] couldn't even I tried to write it down [14:51] one time and the list got pretty long [14:53] and I was like, this isn't helping me [14:54] move forward. I'm just going to I'm [14:56] going to keep focus on what I need to do [14:57] in the future. I mean, to your point on [14:59] vandalism, it's pretty sad because these [15:01] individuals are getting kind of um [15:05] amped up from on TikTok and on Instagram [15:08] to go take action. And I don't know if [15:11] they realize they're committing a [15:12] felony, [15:13] >> you know, and and I saw this individual [15:15] who just got got arrested in I think it [15:18] was either Ohio or Illinois. And Jason, [15:21] he he's he's facing like 15 to 20 years [15:23] in jail. [15:25] >> And I'm thinking I I feel horrible, [15:27] right? It's like I I don't think he I [15:29] hope maybe he realized it, but what if [15:31] he didn't? And and the person who's on [15:33] TikTok or Instagram that's pushing this, [15:35] they're not they're not the ones cutting [15:36] these things down. They're not the ones. [15:37] They're just they're they're doing this [15:39] for likes and views and you know that I [15:42] I don't like. Right. Because when that [15:44] camera is cut down and there's that [15:46] Amber Alert, it's not going to work. [15:49] >> Yeah. And and that and that, you know, I [15:51] still don't have a lot of space in my [15:53] heart for people who choose to commit a [15:55] felony versus go to city council and and [15:58] scream and fight for what they believe [16:00] in. That that is [16:01] >> they join. They could join city council. [16:03] And there have been uh a number of [16:07] cities that have decided to take flock [16:10] down. Am I am I correct? And the [16:12] backlash has caused that. And you're [16:13] okay with that if that's the community's [16:15] choice. I would assume. [16:16] >> Yeah. Yeah. And it's it's obviously [16:18] disheartening when a city chooses to be [16:19] less safe. You know, that that's tough [16:21] to to hear when there's so many other [16:23] options to reduce the efficacy of the [16:26] product versus just take them out. Um, [16:29] you know, to me, the individuals who [16:31] only look at privacy or only look at [16:33] safety are both on the wrong side of the [16:35] equation. Like, we've got to find a [16:37] common ground. That's the the point of [16:39] the democratic process is common ground [16:41] on complicated problems. And I think [16:43] safety is up there on Marlo's hierarchy [16:46] of like things we have to do for our [16:48] community. [16:49] >> It's one of the most fundamental things [16:50] that I believe our government owes us is [16:53] an equal level of safety. And today, [16:56] depending on the color of your skin, the [16:58] and your wealth level and your race, [17:00] sorry, and your sex determines how safe [17:02] you feel. Women feel less safe than men. [17:05] That's a fact. [17:06] >> Sure. [17:06] >> But minorities feel less safe than you [17:08] and me. That's a fact. If you live in a [17:11] a less affluent city and a less affluent [17:13] part of a city, you feel less safe. And [17:16] I think our government has an obligation [17:18] to make us all safe. And that's what [17:20] we're trying to do. And you'll notice [17:23] >> what are the what are the privacy folks? [17:25] Where do you find common ground with [17:27] them? [17:28] >> And where do you feel obviously [17:30] vandalism is an easy one. Okay, [17:31] >> that's an easy one. Yeah, [17:32] >> that's an easy one. But there's a nuance [17:34] here like, hey, I didn't opt into this. [17:37] I don't want my license plate in there. [17:38] Is there an ability for them to opt out? [17:40] And [17:42] >> what is the max um now for the the [17:46] amount that's saved? And then really, [17:48] this is a very nuance point. What's the [17:50] audit trail here? If there's a dirty cop [17:52] as described who's into Tony Soprano for [17:55] a hundred dimes for his betting on NFL [17:57] games and he wants to give the mob or [18:01] somebody like a bunch of license plate [18:03] data from flock, what's to stop the [18:05] dirty come? What's to stop [18:07] >> this and do false matter? But also like [18:10] >> yeah, [18:11] >> what's the largest window people can use [18:13] today on Flock? [18:15] >> Let's let's let's hit on the abuse thing [18:16] because I think that's something that [18:18] you know I was called naive. I I would [18:20] say maybe optimistic. You know, I grew [18:22] up in Atlanta. I trusted the police. [18:24] I've always trusted the Atlanta Police [18:26] Department. And you know, over the last [18:28] year, I've spent a lot of time talking [18:29] to people who don't like Flock. [18:32] >> And I I finally got to the bottom of it [18:34] because I was like, why do you why do [18:36] you hate this company so much? We sell a [18:38] camera. [18:38] >> Um and I realized this woman was like, [18:41] Garrett, I I don't trust the police at [18:44] all. And I was like, ah, that that's the [18:46] root issue. And I was like, but why? And [18:49] she was like, "They lie, they cheat, [18:51] they steal. I don't trust any of them." [18:52] And I'm like, "Okay, man. What if you're [18:55] what if you're right, I'm wrong." Um, [18:57] and so, [18:58] >> and that might be the case for her. [19:01] >> Yeah, she's right. There have been many [19:02] cases of police departments that have [19:05] subverted, you know, even the wearable [19:08] cameras. They turn them off. I mean, we [19:10] have ICE agents wearing masks and no [19:12] numbers. So, the public, [19:13] >> this is rightfully concerned. [19:15] >> It's rightfully concerned. and they're [19:17] bundling you as their partner. Now I get [19:19] it. [19:20] >> Now that's where they have a valid [19:21] concern. [19:22] >> Yes. And so we said, you know, it's [19:24] interesting and I think if you make it a [19:26] more macro topic, anyone in a position [19:29] of power, there's a higher likelihood of [19:31] abuse. Look at CEOs, look at elected [19:33] officials, look at police. When you're [19:35] in power, abuse tends to follow not too [19:37] far behind. And so when we met, you [19:40] know, six or seven years ago, we had the [19:41] audit log. And my assumption was like, [19:43] this is good enough. This is what most [19:45] police systems do is they have this [19:47] audit log. But what I didn't realize is [19:49] he was meeting with a large agency in [19:50] Southern California and they've got [19:52] something like a thousand officers a day [19:53] that use Flock and imagine they're doing [19:56] five or six searches a day. [19:58] >> And imagine on once a month someone one [20:00] person is supposed to review that log. [20:03] >> Yeah. So they're going to spot check 10 [20:05] or 10 or 20 searches which is the best [20:07] they can do and then move on. And so we [20:09] built four months ago, we built this new [20:10] tool called audit assistance because, [20:12] you know, we solved a million crimes [20:14] last year. So we have a pretty good [20:15] pretty good sense for what normal usage [20:17] looks like. And then we had about, you [20:20] know, just under a dozen abuses over the [20:23] last eight years that we had detected [20:25] and said, but we also kind of know what [20:26] abuse looks like. [20:28] >> Tell us what does abuse look like? [20:29] >> So here's a good example. Um, this is [20:32] the easiest one. Let's say you're an [20:34] officer and you'll say you're trying to [20:37] stalk me. You would look for me. You'd [20:39] say, "I'm looking for Garrett's car. [20:40] It's ABC124." And you'd search for it [20:42] and kind of figure out where it is. And [20:44] then the next day, you'd search for me [20:45] again. And then the next day, you'd [20:46] search for me again. That's very [20:48] abnormal. If you were truly trying to [20:50] find me because I committed a crime, you [20:53] would have put my tag on the hot list so [20:55] that anyone else in the police [20:56] department would have been also notified [20:59] because you might be asleep, you might [21:00] be off shift because I'm a criminal. So, [21:02] everyone has a right. Everyone's trying [21:04] to find me. If you're stalking me, you [21:06] don't want anyone else to know what's [21:07] happening. [21:08] And so that's maybe that's the easiest [21:09] one we can flag. But so we built this [21:11] tool four months ago. We kind of put it [21:13] out there not knowing is it going to [21:15] work well? How how how how good is it? [21:18] How much signal noise? And so you've [21:20] seen the headlines over the last few [21:21] weeks. We've caught a lot of bad cops [21:25] and and a ton. [21:27] >> Yeah. [21:27] >> Yeah. And it's it's a ton. It's it's [21:29] more than I ever would hoped. Um, and [21:31] you look at my home state here in [21:32] Georgia where, you know, some of our [21:34] customers have been with us for years. [21:36] And I have to give those chiefs an [21:38] applaud because they're not hiding [21:39] behind it. They're publicly saying, "I [21:42] fired nine officers today, which is [21:44] which is the the most severe offense in [21:46] Georgia for for abusing the system. [21:47] They're out of my police department. I [21:49] don't want them." [21:49] >> Wow. [21:50] >> And I think that's good. And like we [21:51] should be we shouldn't be criticizing [21:53] those guys. We should be saying thank [21:55] you. And who's next? Like who else wants [21:57] to have a clean police department? And [21:58] so we we now are requiring that as a [22:01] mandatory. It's not even a default. It's [22:02] a mandatory feature. You have to have [22:04] turned on. We have to find these bad [22:06] cops and get rid of them. They shouldn't [22:09] have a badge. Amazing. And so this is [22:12] very interesting, I think, because if we [22:15] assume that, you know, let's just pick a [22:17] number, 1% of cops do bad things. Some [22:21] people might think it's all cops, some [22:22] people might think it's 10%. [22:24] >> I'm just gonna say 1%. It it's a very [22:28] small percentage, but as a in a in a [22:31] city with 30 or 40,000 police officers [22:34] like New York City has, I believe, and I [22:36] think Los Angeles 10 15,000 cops, [22:38] >> 1% all of a sudden becomes 100 or 300. [22:42] That's a lot of bad cops. And you have [22:45] set in a way a honeypot for those bad [22:48] cops. [22:49] >> And you know, they probably don't know [22:52] that this is being tracked. My question [22:55] to you is when I make a request, I'm a [22:58] rank and file cop and I go do this. Why [23:01] should rank and file cops be able to do [23:02] this? Shouldn't there be like a higher [23:05] level person who they put the request [23:07] in? Or maybe there could be like a [23:09] supervisor and that supervisor. So it [23:11] says, I'm looking into this case. I [23:13] explain why I'm looking into it. And [23:15] then a senior person like a sergeant on [23:18] duty. They have to say okay, you can run [23:21] that search. And then a third person has [23:24] to look at the sergeants okays and they [23:26] put a note on it. You could also build [23:27] that into the system which is [23:29] >> you know I think like a double key [23:31] system [23:32] >> where two people have to say okay to [23:34] this which now you're a dirty cop you [23:36] got to find a dirty [23:37] >> job. Yeah. [23:39] >> No I think what you're describing um we [23:42] have I can show you the screenshots of [23:44] what it would look like and how it would [23:46] work and it's definitely something [23:47] that's on the whiteboard for how we [23:49] continue to curb abuse. I think the [23:51] challenge we face um and you you've [23:54] dealt with this in your software [23:55] companies. We have some customers who [23:58] have thousands of employees and we have [24:00] some who I was just on the phone with [24:01] the police chief in Texas with 20 [24:03] employees. [24:05] >> The burden of what we you and I just [24:07] described would be pretty pretty big for [24:09] him. Um with just 20 officers, he's also [24:12] not using the product every day. He [24:13] doesn't need to. There's only count of [24:14] 2,000 people. [24:15] >> So the audit log is very slim. [24:16] >> Audit log is slim. So he doesn't need [24:18] any of the things we're talking about. [24:19] And so I do think this this is where [24:21] defaults do matter and this might get to [24:23] a level of nuance where okay for big [24:25] cities this is these are the rules for [24:27] medium cities these are the rules and [24:28] for small towns you have different rules [24:32] no different than theirmies to become a [24:34] police police officer in New York City [24:36] is harder it's a longer academy than to [24:38] be a police officer in a small town [24:40] because there's just different [24:40] expectations so yeah you're spot on [24:42] we're not done like I think abuse is a [24:44] problem that's industrywide it's not [24:47] just a flock thing [24:48] >> yeah And you look at like California, [24:50] you know, there was 7,000 reported [24:52] abuses of their state database last [24:54] year. 7,000 [24:55] >> 7,000 of the state database. Yeah. Which [24:57] means you should have some percentage of [24:59] that in your system. And I saw the [25:01] Washington Post did a story on this. And [25:03] so the company's evolving and reacting [25:06] to that. And so you are responsible now [25:09] as a technology provider with the worst [25:11] behaviors of your customers. And so the [25:14] lesson there is even if you want to say [25:17] hey we just sell a tool the reality of [25:19] running this company since it is uh you [25:22] know people have to give up some level [25:24] of privacy to get that security you [25:27] actually have to own it there's just no [25:28] choice and I think previously your [25:30] approach was hey it's a tool do with it [25:32] what you will you made that choice [25:35] >> which is not exactly an unreasonable [25:37] thing to say people sell laptops and [25:39] cameras and guns to a local police [25:42] department they send tasers to a local [25:43] police department. We don't say to the [25:45] taser company, "Hey, you're responsible [25:47] when somebody gets tased." And it was an [25:49] inappropriate tase. It was an [25:51] unjustified [25:52] >> taser. But here we are. This is [25:55] different. This is qualitatively [25:56] different because America and Americans [25:59] really value privacy. [26:05] >> Sales tax rules are changing. Starting [26:07] January 1st, California will tax certain [26:10] software for the first time. Numemeral [26:12] can help. Numemeral is the sales tax [26:15] solution for fast growing businesses [26:17] handling registrations, filings, tax [26:19] rates and compliance across the US and [26:22] over 70 countries. So you can stay ahead [26:25] of risk with NL's guarantee. Sell with [26:28] confidence. You grow the business. [26:30] Numero handles the sales tax. Learn more [26:33] at numeral.com. [26:38] to your point, our our I think our view, [26:41] my view of our responsibility has [26:43] evolved. [26:44] >> Um, now I would prefer if every elected [26:47] official understood how all this [26:48] technology worked perfectly [26:50] >> and made decisions that represented [26:51] their community values. I don't think [26:53] that's a reasonable expectation because [26:55] that means they also need to know [26:56] everything about data centers. They need [26:58] to know everything about water works. [27:00] like they're relying on their police [27:01] chief [27:02] >> to tell them what is best [27:04] >> and that chief's relying on me to [27:06] educate him on what I think is best. [27:08] >> And so I do think the responsib [27:10] responsibility falls on us to give those [27:12] strong defaults and then allow them to [27:15] pivot against it. Um and still respect [27:17] the democratic process but focus on the [27:19] defaults. [27:20] >> Yeah. And as the lead candidate, you're [27:23] going to take the arrows. But there's [27:24] other people. My friend Jamie Simonoff, [27:26] who's an incredible entrepreneur, [27:27] >> great entrepreneur [27:28] >> and a dear friend of mine for many [27:29] decades. He has Ring famously, and [27:32] during the Super Bowl, he did this thing [27:34] like, "Hey, with Ring cameras, and I [27:35] guess I don't know if they're considered [27:36] your number one competitor because they [27:38] sell to individuals and then allow the [27:41] data to be used by police if people opt [27:43] into that, which is an individual's [27:45] right." [27:46] >> They were like, "Hey, we can find a lost [27:47] dog." And and people lost their mind [27:49] over that because they were like, "Hey, [27:51] >> if you can find a lost dog, then you [27:54] can, you know, have somebody stalking [27:56] what whatever people's valid concerns [27:58] are." And they had to turn that feature [28:01] off is my understanding after releasing [28:04] it. [28:04] >> Yeah. Know, I've I've I've known Jamie [28:06] for a long time as well. I don't think [28:08] they're a competitor. I'd say they're a [28:09] friendly. We have the same mission, [28:11] which is to build safer communities in [28:13] America. and and I think him and I have [28:15] learned a lot over the last year of you [28:16] know maybe underestimating people's [28:18] concerns with how technology could [28:21] impact them um and needing to maybe go [28:23] slower in some cases like we you know we [28:25] haven't gotten into the AI topic but [28:26] it's an area where I think we're doing [28:28] it smarter now because there's a version [28:30] of flock where we would have just ripped [28:32] and launched AI feature after AI feature [28:34] and said this is awesome and we're going [28:37] intentionally really slow [28:39] what is the opportunity and the concern [28:42] with AI give us the practical, you know, [28:46] um, you know, in if we were in a police [28:48] state, if it was China and they're using [28:51] this like the CCP has a a literal police [28:55] state, that's very triggering for [28:57] Americans because they see that then [28:58] they see flock on this end and ring on [29:00] this end and they say, "Hey, what's the [29:02] distance between these two?" [29:04] >> And AI's the obvious one because they're [29:06] tracking people with facial recognition [29:08] and they're watching their behavior. [29:10] They're like, "This person littered. [29:12] this person raise their voice like it's [29:14] >> pretty gnarly. Yeah, it's very [29:16] different. So tell us about the AI [29:17] features and then what the detective [29:20] wants most because the detective wants [29:22] to close cases [29:24] >> reasonably [29:25] >> and then what you're saying like o gosh [29:27] that would close the case but maybe it's [29:29] too much of a privacy concern. [29:30] >> Yeah. I mean let's talk about a simple [29:32] one that's kind of in debate right now [29:34] which is 911 systems. Um, today, you [29:37] know, when you call Delta, maybe you get [29:39] an AI agent, maybe you get a human, but [29:42] you're booking a flight, so the risk of [29:44] kind of hallucination is pretty low. [29:46] Maybe they book you on the 7 p.m. [29:48] instead of the 4pm, and you're like, [29:49] "What just happened?" But it it's fine. [29:52] >> Imagine someone broke into your home and [29:54] you call 911 and we've now switched over [29:56] our 911 system to an AI bot and it [29:59] starts talking to you about what you had [30:01] for lunch. [30:01] >> Yeah. No good. [30:02] >> No good. and and the risk the risk is so [30:05] high and we have no idea how these [30:08] things work and there's no good way to [30:10] control them. That's a I think that's [30:12] pretty well in sap. And so you see [30:13] cities due to budgetary reasons and due [30:15] to staffing issues kind of leaning in [30:18] pretty aggressively and there's been [30:20] some bad cases where the AI's gotten it [30:23] really really wrong. [30:26] >> And so I think there's a smarter way to [30:27] do it. Like the smarter way to do it it [30:29] would be okay look we're going to we're [30:30] going to have AI but AI is only going to [30:32] pick up when the phone lines are fully [30:34] busy. [30:35] >> Yes. Okay. [30:36] >> So it's like okay that's smarter but [30:37] that's not what's happening. like [30:38] there's a reckless way to do it which is [30:40] just flip a switch or maybe you start [30:41] with just [30:42] >> do you handle 911 systems? [30:44] >> We don't the ind the indust we don't do [30:46] 911 calls but I'm watching this unfold [30:49] and I'm going man this is a like this is [30:51] a huge risk. Um I look for [30:54] >> I suppose you could use AI to look at [30:57] patterns and say these are abnormal [31:00] patterns for this car. This car is [31:02] slowly driving up to this house, then [31:05] driving away, slowly drove up to the [31:07] next house, drove away, and it did that [31:09] six times. [31:10] >> Is this somebody delivering a lot of [31:12] burritos, or is this somebody casing the [31:14] joint? Like, [31:15] >> that would be what the detective would [31:17] be like. Hey, yeah, let's get pregs [31:19] here, precognition from Minority Report [31:21] to predict a crime. [31:24] >> Yeah. [31:24] >> Because somebody wants to join. Yeah. [31:26] >> Yeah. That I don't want to do. Like I [31:27] think we've been really we've been [31:29] really consistent that um we never want [31:31] to be a company that defines suspicion. [31:34] >> Um that's really dangerous. [31:36] >> Yeah. [31:36] >> Oh yeah. But it's it's a it's an [31:38] interesting kind of like happy hour [31:40] topic of like what does that world look [31:42] like? [31:43] >> But for us I think there's it's really [31:45] valuable that there's a human in the [31:46] loop when it comes to public safety. [31:48] We're talking about people's lives. [31:49] We're talking about pe people's like [31:51] it's it's really important stuff. So [31:53] >> we always want to have a human in the [31:54] loop. I think there is some some [31:56] products that we could build but the [31:58] difference is like it you know looking [32:01] at knowing the cost of getting it wrong. [32:05] You have to deploy this with third party [32:09] proper attestation that it's that it's [32:12] working as expected. It shouldn't be up [32:14] to me. It shouldn't be up to my [32:15] customers. Someone who actually doesn't [32:17] want this to work should come in and [32:19] say, "Well, guess what? It it does [32:21] actually like this is actually really [32:22] good and it's making officers more [32:24] effective. It's making officers more [32:26] objective, less subjective. Like that [32:29] should be our end goal is that AI is [32:30] used in a way that increases the number [32:33] of crimes they can solve and decreases [32:35] the likelihood if they get the wrong [32:36] person. That that's an easy ideal state. [32:38] And I just don't think we should rush [32:40] there. I think we should be really [32:41] deliberate about how we get there. [32:42] You've lost, I think, dozens uh of [32:45] cities of the 6,000. So, some cities are [32:49] now saying, "Hey, this isn't worth the [32:51] trouble." Or too many people protested. [32:54] Take me through what's happening in the [32:55] last year with those I think it's 60 or [32:58] 70 cities maybe out of 6,000 if I got my [33:01] numbers correct. 1% have said, "Okay, [33:03] this isn't for us." [33:04] >> Yeah, it's tough, right? Because uh I [33:07] want everyone to be safe. That's why I [33:08] started the company. Um, and I know what [33:11] they've had and I know the results we've [33:13] had. And unfortunately, the majority of [33:16] the time they're making a decision off [33:18] of misinformation. Um, I remember [33:20] vividly the one of the first customers [33:22] that chose to walk away did so because [33:25] of facial recognition. They went online [33:27] and googled ALPR facial recognition. [33:30] They saw one of our competitors website [33:31] and said, "We're getting rid of this [33:33] flock thing." [33:34] >> They didn't realize they were looking at [33:35] the wrong person's website. [33:37] >> Yeah. And that's tough because like [33:39] that's on me, right? That's on me to [33:40] better communicate to to these decision [33:42] makers that like here's what we do, [33:44] here's what we don't do. Um that's been [33:46] the bulk of it. I think the other thing [33:48] that uh has been tough to see, Jason, is [33:52] >> I saw this one city in in Washington [33:55] just just it's so sad. They turn the [33:59] cameras off. The next weekend, a little [34:02] boy gets shot, [34:05] car drives off. It's gonna be a cold [34:08] case. And luckily, the city next door [34:10] still had Flock cameras, found the car, [34:13] made the arrest, and look the that the [34:17] kid that didn't live, which is horrible. [34:20] >> At least their family knows justice was [34:21] served. [34:22] >> And so, I think we're making really [34:24] consequential decisions on on inaccurate [34:26] information, and that's tough. And I [34:28] just we have to do a better job as a [34:30] company like communicating to our [34:32] customers and their customers customers [34:34] which would be the community about what [34:36] this technology does and what it doesn't [34:37] do. And and for me at least when I when [34:39] I speak to let's call it 99 people out [34:41] of 100 and explain the facts they go [34:44] like why is this a debate? [34:46] >> This feels like we have control as a [34:48] city. We can set limits. We can set [34:49] guard rails. I want to be safer than I [34:52] feel today. I've never met someone who's [34:54] like I want to be less safe. I predict [34:55] as you explain this and you change the [34:59] defaults and the audits get in place and [35:01] the dirty cops get, you know, taken out [35:04] of this that you'll see some number of [35:05] those cities, you know, and the city [35:09] councils readress this when things cool [35:12] down a little bit. And, you know, this [35:13] happened here in, you know, where I live [35:15] now in Austin, Texas. Austin canled [35:18] flock or didn't renew whatever in 2025 [35:22] according to my notes. Then we we you [35:24] know people have guns here. I I have a [35:26] gun. I carry a gun. Everybody's got [35:28] guns. And so we had this uh crazy uh [35:31] shooting with a bunch of kids. Uh and it [35:34] was a similar situation to what you're [35:36] describing. They were committing these [35:37] crimes in Austin proper. [35:40] They drove out of Austin and Manor [35:43] Police Department did have flock. They [35:45] picked it up and 28 hours later the [35:48] three teens were arrested and they were [35:51] like juveniles 15, 16 and 17 here [35:53] according to what I'm seeing. And that [35:56] was the wakeup call here and that [35:58] brought it back to the front and now [36:01] everybody's debating it again in Austin. [36:04] Okay, the ACL ACLU wants 48 hours I [36:07] think is their proposal. Okay, that's [36:09] the ACLU. [36:10] >> That's reason. Yeah, that's their that's [36:12] a reasonable stance. This is the privacy [36:14] organizations, privacy organizations, [36:15] civil liberties. We get it. [36:17] >> You know, most cops probably want 30 [36:20] days. Communities are going to set it at [36:21] seven or 14 days. That's it. You have [36:24] competitors. Tell me about like the [36:27] competitors approach here and how you [36:28] deal with that as an entrepreneur. [36:30] >> So, are they just saying, "Hey, we're [36:32] just going to sell you the technology [36:33] and we're not going to and we're just [36:34] going to keep our heads down and we'll [36:35] be the uh do they go in and sell [36:37] themselves as a we're flock without [36:39] limits?" So to your earlier point, yeah, [36:40] we've had close to 20 cities already [36:43] turn their cameras back on. [36:44] >> Oh, interesting. Yeah, [36:45] >> because they realize like they are now [36:47] less safe and they have an obligation to [36:49] their constituents [36:51] um to to be safer. And it is it is it is [36:53] tough as an entrepreneur because you [36:55] want to you want to build as big of a [36:57] business as possible, right? But you [36:58] also want to do it in a way that when [37:00] it's all said and done, your kids go, [37:02] I'm so proud to be that guy's son or [37:06] daughter. So you have to make decisions [37:08] that you you you you're going to be [37:10] comfortable with 10 years later. And I'd [37:12] rather lose, you know, I had a lot of [37:14] concerns about requiring, you know, [37:17] accountability for all of our customers, [37:18] knowing very well that I was going to, [37:20] you know, upset a lot of customers [37:22] potentially. And I said to myself, look, [37:24] if a customer doesn't want to be held [37:26] accountable, I don't think I want them [37:27] as a customer. And they can go to one of [37:29] my competitors who doesn't have these [37:31] tools, doesn't have these requirements, [37:32] and that's okay with me because this is [37:35] just business, right? I it's it's just a [37:37] business. Um now it's a business I think [37:39] serving a very important mission but I I [37:42] want to sleep at night and I do today [37:44] and our competitors to your point yes [37:46] they run more reckless they run without [37:49] these tools without the scrutiny and I [37:51] think it's going to catch up to them. It [37:53] will catch up to them eventually. [37:54] >> Yeah. I mean as you are willing to lose [37:58] customers who want the un you know [38:01] tethered system without you know proper [38:04] uh constraints. Yeah. You'll lose some [38:06] of them, but I don't think you want them [38:07] as customers anyway because it's going [38:08] to blow back on you. San Francisco [38:10] famously with these ALPRs. Automated [38:13] license plate something. Automated [38:15] license plate reader. Yeah. [38:17] >> So, the ALPRs you said earlier, you're [38:20] also keeping if there were dents in the [38:22] car or the color of the car or like a [38:24] the make and model of the car. Is that [38:26] right? [38:26] >> Yeah. So, it's pretty it's pretty cool [38:28] technology and pretty effective because [38:30] if you think about what happens on a 911 [38:32] call, most people don't instinctively [38:34] look at the license plate and remember [38:35] it. They see the thing that was unique [38:37] about the car. Oh, it had it had a [38:40] damage on the right side. I remember so [38:42] vividly we had this homicide case near [38:45] me in Atlanta and all the person could [38:47] remember was that the back right tire [38:49] was missing a hubcap. M [38:52] >> and so the detectives would go in and [38:54] say, "Okay, show me all the cars in this [38:55] 15-minute window that did not have a [38:57] backright hubcap." And that's the [38:58] beginning of an investigation. It's not [38:59] the end of an investigation. That's just [39:01] the the way to get started. And that's [39:03] all they had. And because of that great [39:05] detective, he's like he's I've got to [39:07] know him pretty well because he's like [39:09] he's very good at his job. He worked his [39:11] butt off and and found the found the [39:13] person. Um but that was the start. And [39:15] so yeah, so he's like San Francisco. And [39:18] I I love I saw a tweet uh the other day [39:20] uh which was you know the flock effect [39:23] and it's like crime is down, stolen [39:25] vehicles is down. And look at we we we [39:28] are one small part of a very big [39:30] transformation that's happened in that [39:32] city including stronger elected [39:34] officials, stronger enforcement of the [39:36] law, a strong police chief, a strong DA, [39:38] a community that demands to be safe. [39:40] Like but that's what that's what should [39:42] happen. If you live in San Francisco, [39:44] you deserve to be safe. It's one of the [39:45] best cities in the country. And San [39:47] Francisco [39:48] has a lot of problems and people are [39:52] very libertarian, privacybased, live and [39:54] let live based. It is kind of the [39:57] problem having lived there is that [39:58] people are so openhearted from the [40:01] summer and it literally comes from the [40:02] summer of love and the fact that it was [40:04] a mining town like they just have this [40:07] you know live and let live attitude. [40:09] Obviously, you know, a lot of the [40:11] culture that's emerged there is, hey, [40:12] let your freak flag fly. Do what you [40:14] want to do. Even to the point of, hey, [40:17] you know, if people are stealing, um, [40:19] why are they stealing? Like, what [40:21] happened in their childhood and and [40:23] people there have had enough. the fact [40:25] that San Francisco um and I and I think [40:28] this is the correlation. Do you have the [40:31] ability and do you have the entitlement [40:35] to say I don't need to worry about this [40:37] because I have security at my house. I [40:39] have a security guard. I have a personal [40:42] driver. I have personal security. Which [40:44] by the way, a lot of rich people have [40:45] now some combination of these things at [40:48] a minimum. Like I just described I have [40:50] cameras, I have people, you know, all [40:52] this stuff. And so, you know, you start [40:54] painting that picture. It's it is you [40:57] you have to be extremely aware of your [41:00] own privilege. [41:01] >> Yeah. [41:01] >> That you don't have to deal with crime. [41:03] And man, my brother was a cop during the [41:05] stop, ask, and frisk era in New York. [41:08] >> The statistics during stop, frisk, and [41:10] ask, which was a violation of the fourth [41:13] amendment, [41:14] >> like you're stopping people based on [41:16] like they're hanging out. So, it [41:18] definitely was a little step too far. [41:21] But New York in the 70s and 80s was [41:23] dangerous as hell. Like there were a lot [41:25] of guns on the street. My brother took a [41:26] couple of guns off the street. Cops were [41:28] in a really tough situation. The people [41:30] in the poorest communities were the most [41:33] in favor of stop ask and frisk. They [41:36] always leave out the ask because they're [41:38] like, "Oh, we're just going to stop you [41:39] and frisk you." No, they would stop and [41:40] say, "Hey, what are you up to?" And uh, [41:42] "Hey, you're hanging out in the corner. [41:43] What's going on?" And you know, they [41:45] just try to get a read on people, which [41:47] by the way, if you ever go to Israel or [41:49] you go to some countries, they ask you [41:51] 20 questions. Why are you here? Hey, uh, [41:53] you know, where you staying? Like, they [41:55] want to get a read on you, whether [41:56] you're going into Israel or Tokyo. They [41:58] might ask you a couple personal [41:59] questions, which would be offensive to [42:02] Americans. But the correlation was [42:04] really people who experience crime are [42:07] in favor of it. People who live in Pack [42:10] Heights, they don't need to worry about [42:12] this, man. something goes down in Pack [42:13] Heights, there's 20 cops and 50 private [42:16] security guards, you know, rallying [42:17] around one homeless person, one vagrant, [42:21] you know, one person who's in a car that [42:23] doesn't look like it should belong [42:24] there. It's like they're going to get [42:26] swarmed in Pack Heights. [42:27] >> Yep. No, I mean, you hit the nail on the [42:29] head and it's not spoken about as much, [42:31] but it's worth looking at who is so [42:33] against what we do and who is so for it. [42:36] And it's typically people who don't have [42:38] any other way to protect themselves, [42:40] who need this, right? who want this. And [42:43] and for those who have a gate and have a [42:45] guard and have cameras and have a [42:47] security system and they have Jaime's [42:49] incredible doorbell camera, right, and [42:50] they have all these products and these [42:52] services, they're like, "Why does why [42:53] does anyone need to feel any safer?" [42:54] It's like, "Because most people don't [42:56] have any of this stuff and they don't [42:58] feel safe when they go to bed." And [43:00] yeah, it's it's it's quite telling. [43:01] >> I'm rooting for you to get everything [43:03] right and to continue making communities [43:06] safer. I'm also rooting for communities [43:09] to make their own decision. 100%. [43:12] >> I think that is the key part of this [43:14] that people need to embrace as well. If [43:17] you have a if you've chosen to live in a [43:20] gated community, you have made that [43:22] choice for security in all likelihood. [43:25] And if poor people or people who don't [43:29] get to live in a gated community and you [43:31] were to ask, you know, people with a [43:32] family, hey, would you like to live in a [43:34] gated community or would you like to [43:35] live with not in a gated community? Like [43:38] a hundred out of a hundred people. I I [43:40] mean maybe it's 95. I think if you [43:42] offered like the average Joe or Jane, [43:46] hey, would you like to live in a would [43:48] you like your community to be gated with [43:49] three gates and there's a security [43:50] guard? They'd be like, heck yes, but I [43:52] can't afford that. How much is that [43:54] going to cost? So, this is where, you [43:57] know, I think people with children, [44:01] people with homes, they they just have a [44:02] different view of this than maybe when [44:04] you're younger and you can afford to [44:06] say, "Listen, I don't have kids and I [44:07] care about my safety. I care about my [44:10] privacy, not my safety. Yeah. And and I [44:12] understand it. Like I I don't want the [44:14] government to have like a national flock [44:17] database. I don't want the government to [44:19] do facial recognition. I find that like [44:21] a bridge too far. [44:22] >> The license plates for two weeks or 30 [44:25] days. I kind of feel okay with it. I'm [44:28] sorry if people are disappointed in me [44:31] for that, but uh as a parent and a you [44:33] know, somebody in a community, I'm okay [44:35] with my community having license plate [44:36] readers for up to 30 days. That's that's [44:38] the decision I came to. [44:39] >> Well, you know, and Jason, I think to [44:40] your point, I I had this interesting [44:42] conversation with um an elected official [44:44] in Illinois who was one of our largest [44:45] antagonists, very anti- flock and really [44:50] sad, really scary. She was the victim of [44:53] political violence. Um someone shot at [44:55] her house and the police used a flot [44:58] camera to find that person [45:00] >> and lo and behold, to her credit, she [45:02] changed. But she did change and she's [45:04] now one of our largest advocates because [45:07] she realizes you don't get to pick when [45:08] you're the victim of violence. You don't [45:10] get to pick. Sometimes it's random. [45:12] Sometimes it's intentional, but everyone [45:14] wants justice [45:16] when something bad happens. And everyone [45:18] wants to know that they're going to be [45:19] safe. And so I think you're right that [45:21] my my hope is that this is these kind of [45:23] conversations actually lead to fruitful [45:24] debates about how we have both um and we [45:28] don't let you necessarily allowed [45:29] minority uh dictate you know us moving [45:32] towards a less safe society. That'd be [45:34] pretty sad. [45:35] >> Yeah. And listen you have been in what I [45:38] would describe as a PR disaster the last [45:40] year over this. How far through the PR [45:44] disaster are you? You've changed a lot [45:45] of the communications. Yeah. [45:47] >> And it is this something Flock can [45:50] recover from or is this like, [45:53] >> you know, company killing? Like, oh my [45:55] god, we're we're screwed. Like, just be [45:57] honest with us about like the PR [45:59] disaster of the last year, how that's [46:01] affected you and how that's informed how [46:04] you, you know, run the company and how [46:07] damaging has it been? You know, I think [46:09] the biggest damage is the internal [46:13] morale, right? Because you go on to X or [46:16] you go on to Reddit and you go, why why [46:18] are people so mad at us when we've been [46:20] doing the same thing for nine years? And [46:22] and why are they so mad at us for things [46:24] we don't do? And we just we just want [46:26] people to be safe. And that's tough, [46:28] right? It's really hard for your your [46:30] your work to be challenged so [46:33] viscerally. [46:34] I think the the interesting part is it's [46:36] brought us closer to our customers, you [46:38] know, because we're because they they [46:40] they have a job to do, which is to keep [46:42] their community safe and they see this [46:43] as one of the the most, if not the most [46:47] critical part of of how they do their [46:49] job. [46:49] >> Like there was a chief who was like this [46:51] is the best thing since DNA. [46:52] >> Yeah. [46:53] >> In terms of fighting crime and I agree [46:54] with them. So that's on the good side. [46:56] And I think you know the business is in [46:58] good shape. You know, we raised we [46:59] raised more money at the beginning of [47:00] the year. Uh company's still growing. We [47:02] could be growing faster if we used to [47:04] have zero churn. Now we have, to your [47:06] point, 1%. [47:08] >> Okay. Well, I would like to be back at [47:09] zero. And I'm going to fight as hard as [47:11] I can to get the right message out [47:13] there. To your point, communicate [47:15] better, listen more. Um, and I think if [47:18] we do that, we're going to be just fine [47:19] because at the end of the day, everyone [47:21] wants to be safe. That hasn't changed. [47:23] And that will probably never change. [47:24] Just like everyone wants to have food, [47:26] everyone wants to have a roof. Uh, those [47:28] are just these core needs. [47:29] >> As Maslo's hierarch needs, safety is up [47:31] there. And it's, you know, I think for [47:34] your company, like you're going to just [47:36] have to lead and you're going to have to [47:38] be ahead of these things. And I, you [47:39] know, I look at Zuckerberg and Facebook, [47:42] >> they've always been playing defense and [47:44] catchup on the impact that, let's say, [47:47] Instagram had on young teens and eating [47:51] disorders and addiction to social media. [47:54] uh X and Twitter had like political, you [47:58] know, nonsense and like people trying to [48:00] subvert democracy. Like all of these [48:02] social networks had a a serious downside [48:05] that's now been quantified, etc. It [48:07] would have been so much better if he was [48:08] ahead of it and said, "Listen, I have [48:10] kids. [48:10] >> I understand where you're coming from. I [48:12] think there should be an age gate." And [48:14] he just wrote that manifesto last week [48:15] and I'll give him credit because I think [48:17] he was looking at it saying there are [48:19] obviously selfish business needs. he [48:21] wants to be part of the AI revolution, [48:22] but I think he was also identifying and [48:25] trying to get ahead of and show some [48:27] empathy for, hey, yeah, you you could [48:29] lose your job because of AI. Here's how [48:31] I think if you use the tools, you [48:33] wouldn't lose your your your job. And if [48:36] you have open source and it's available [48:37] to everybody, not just the rich, and we [48:40] give you agents and free AI that other [48:43] people are paying thousands of dollars a [48:44] month for, hey, maybe that levels the [48:46] playing field. And I think that's what [48:47] CEOs have to do right now in this media [48:50] saturated. [48:51] >> You know, a story can just go viral so [48:55] fast and you got to just get ahead of it [48:56] and anticipate these things and own them [48:58] and just be frank about it. Tucker [49:00] Carlson demolished you last week. Have [49:02] you have you reached out to Tucker to go [49:04] on his pod? [49:05] >> No, I don't you know I don't know if if [49:07] if that's something that's in my my deck [49:09] of cards, but uh no, I appreciate it. I [49:11] I like that. Uh you know, I do like [49:13] t-shirts, I guess, is what I'd say. Uh I [49:15] wear t-shirts. I I got I put a polo on [49:17] for you. I thought a collar would be [49:19] appropriate, but [49:20] >> I wear a polo, too, in the hot. [49:22] >> Yeah, you look you look great. Yeah, you [49:23] look great. But no, I think look, we we [49:25] want to we were on the Today Show this [49:26] morning. We're on Good Morning America. [49:28] We want to go to places where the [49:29] community listens and we want to have [49:31] meaningful conversations and uh yeah, [49:34] we're not done. We got a we got a long [49:35] road ahead of us and I think, you know, [49:37] we didn't get to drones yet. [49:39] >> So, how did they Yeah, that's how they [49:42] drones. So drones is probably our [49:43] fastest growing business unit right now. [49:46] You know, our LPR business is just under [49:47] 50% kind of of our forward-looking [49:49] revenue. Um, so it's a big part of our [49:52] business, but we've got drones, we've [49:54] got live video cameras, we've got some [49:55] really cool software. And you know, I [49:57] think the story I share with you about [49:59] drones, which is one of my favorite [50:00] stories, is so these drones live um in a [50:03] kind of a HVAC control box, a dock, per [50:06] se, on top of a roof, typically like a [50:08] police precinct or a fire department. [50:10] And when a 911 call happens, instead of [50:12] sending an officer first, you can send a [50:14] drone. And so there's this really [50:16] interesting case not too far from you up [50:17] in the Dallas Fort Worth area where a [50:20] woman, you know, at late at night at a [50:21] gas station called 911 and said, [50:23] "There's a man walking back and forth [50:25] with a gun. I don't feel safe." Um, [50:27] which obviously it's not illegal to to [50:29] garnish a weapon in the state of Texas, [50:32] but she was scared. Right. And [50:33] >> well, you can wear one, but you can't be [50:35] waving it around. [50:36] >> Yeah. And he was fair enough. Fair [50:37] enough. Yeah. And and so she called 911. [50:40] And normally what would happen, right, [50:41] is like one cop would have heard that, [50:44] flip on the blue lights, race down [50:45] there, their hearts racing, he's got his [50:47] gun ready, and someone's about to die. [50:50] >> Instead, they launch the drone. The [50:51] drone's at 400 feet, flying 60 miles an [50:54] hour, so it gets there in under a minute [50:56] >> as the crow flies. [50:57] >> Yep. And it's got a 40x optical zoom, so [50:59] it zooms all the way in. [51:02] It's not a gun. It's a lighter. It's one [51:05] of those toy like lighter things. [51:07] >> Yeah. [51:08] >> That call the woman back and say, "Hey, [51:09] ma'am, it's just a lighter. You're [51:10] okay." And I share that story because so [51:12] much of my businesses have related [51:14] around solving crime, but it's more than [51:15] crime. It's about safety. It's like, [51:17] "How do we help the community feel [51:18] safe?" And this is an example where like [51:20] we demitated, we mitigated what would [51:22] have been a really bad situation for [51:24] everyone. And we have thousands and [51:26] thousands of stories like that where [51:28] communities are safer now with our [51:29] drones. They're safer with our cameras. [51:31] And to your earlier point, we just got [51:32] to do a better job communicating how it [51:34] all works. And we got to teach our [51:35] customers how to do that, too. Um [51:37] because we're only one company. We got [51:39] thousands and thousands of customers. We [51:40] we need their help too and telling the [51:42] story. [51:42] >> Yeah. I think that the new tool you're [51:44] talking about with the audit trail, um I [51:46] think that has great potential. I think [51:48] a third-party auditor, so if you sell [51:50] the system, it can come with thirdparty [51:53] auditing. You know, people use Ernston [51:54] Young or whatever to audit books. What's [51:56] the equivalent here of an auditor that [51:58] would say, "Hey, the drone got launched. [52:01] this isn't a dirty cop sending a drone [52:05] to stalk their ex you know spouse uh or [52:09] you know break somebody's privacy and [52:11] we're just going to review in real time [52:13] slash whatever every drone flight [52:16] because that's now you start sending [52:17] drones Tucker's going to look at that or [52:19] ACLU is going to look at that and say [52:21] okay so now you're like peeking in my [52:23] window so there's got to be massive [52:25] safeguards on the drone getting deployed [52:27] is it only deployed if there's a 911 [52:29] call is that the [52:30] >> this is the next big like chapter for [52:33] law enforcement is is doing a better job [52:36] implementing drones than they've done [52:38] with historical products because you [52:39] look at something like body cameras and [52:41] I'm not sure how much you follow this [52:42] there was massive outrage on body [52:44] cameras and now it's like wait a second [52:45] this actually really good it's a better [52:47] tool and and same with non-lethal [52:48] weapons um and so no we we there's still [52:51] more work to do there and to your point [52:52] on the third party auditors you know one [52:54] you've got a city manager [52:56] >> that is effectively his or her job is to [52:58] hold the police chief accountable [53:00] >> and they can the audit logs. They can go [53:02] in there themselves. [53:02] >> They can see all the logs. Yeah. They [53:04] can see the audit logs. They can also [53:05] see the flight patterns. So, most of our [53:06] customers have a transparency portal. It [53:09] has the audit log. It has how many [53:10] cameras they have. It has their drones. [53:12] It has their drone flights. Now, there's [53:13] a lag, right? We want a delay. Um, [53:18] chasing to the crime, [53:19] >> an active criminal investigation. Yeah. [53:21] So there's a lag, but I I look at that [53:22] in and what's going to be hard for our [53:24] customers is getting comfortable with [53:27] this ever growing expectation of both [53:29] transparency and accountability. [53:31] >> Yeah. [53:31] >> And we're going to push them. We're [53:32] going to push them hard, but the [53:33] community is also going to push them as [53:34] hard. [53:35] >> And some of these things are like 911 [53:37] calls in many jurisdictions are public [53:40] >> public record. [53:41] >> They're public record. And same thing [53:43] with uh radios, although some places [53:46] don't have those as public record. They [53:47] don't have the they have the radios [53:49] encrypted. So drone flights is one where [53:51] like you know should drone flights just [53:53] be on the website like here's all the [53:55] drone flights that occurred and here's [53:57] the 9/11 calls that are with them or [53:58] people doing that or your does your [54:00] software do that for them? [54:01] >> We it would be possible. We've never had [54:04] a customer asked to do it. We have this [54:05] concept of like a spectator mode which [54:08] would be available for, you know, city [54:09] officials and and command staff. But I [54:11] think that your question is the right [54:12] one. And [54:14] >> I think as an industry, you know, we're [54:16] going to have to make these decisions [54:18] sooner than later of what that default [54:20] is because we have a default now, but [54:22] there's very little state regulation. [54:24] There's zero federal regulation outside [54:26] of air safety, which is important, but [54:28] but nothing yet on state regulation or [54:31] city regulation for the [54:34] >> Yeah. So the short so you've got you got [54:36] New Hampshire at three minutes [54:38] >> and then [54:39] >> three minutes for license plate reading. [54:41] >> Yeah. [54:41] >> Okay. That's [54:42] >> that's New Hampshire. That's New [54:43] Hampshire though. So die. Okay. Got it. [54:46] >> And and then the next New York state is [54:47] 21 days. [54:49] >> Interesting. [54:50] >> That's the state of Washington. So [54:51] >> Okay. Well, here here you are folks. [54:53] Like if you want to really live off the [54:56] grid in privacy, [54:57] >> New Hampshire. Live the live free or [54:59] die. Garrett, I appreciate your [55:01] candidness and um good luck being in [55:03] such a controversial space and getting [55:05] ahead of these things. I wish you great [55:07] luck with it and uh we can agree, I [55:09] think, publicly for everybody to agree [55:11] and disagree on the margins about these [55:13] things, but it's up to the community to [55:15] make these decisions. I think and I [55:17] think that's the high order bit here. [55:19] And I think that's what you're starting [55:21] to really get right is, hey, this is [55:23] your community choice. You don't want [55:24] it, turn it off. You want it for 48 [55:26] hours, put it at 48 hours. We're going [55:28] to have some reasonable caps on this. I [55:30] wish you great luck with it and uh thank [55:32] you to everybody for tuning in to the [55:34] All-In interview program. Bye-bye. [55:36] >> I'm going [55:53] all in.