Insurers Carve Out AI Coverage as Robot Injury Suits Threaten Payouts

The History of Insurance and the Coming Age of Humanoid Robots With Clanker Labs — Ai4

Ai4

As carriers drop silent coverage to avoid massive accident liabilities, specialized underwriting checklists will become the de facto safety standard governing commercial robot deployment.

AI-assisted summary of The History of Insurance and the Coming Age of Humanoid Robots With Clanker Labs — original source. Check the original for full context.

Source published:

Detailed notes

## Background Published on YouTube by the Ai4 channel on August 15, 2026, this presentation features Darien Acosta, CEO and co-founder of Clanker Labs. According to the video description, the talk traces a 160-year historical pattern of how insurance converts uninsurable emerging machines into regulated safety standards, forecasting the coming liability reckoning and timeline for humanoid robots and physical AI.

## TL;DR Darien Acosta illustrates how major technological shifts—from 19th-century steam boilers and electrification to nuclear power and cyber risk—rely on insurance underwriters to establish mandatory safety codes when machines fail. He outlines the 'five stages of insurance grief,' showing how liability moves from silent coverage and knee-jerk exclusions to specialized underwriting and eventual societal normalization. Applying this framework to the modern era, Acosta highlights recent chatbot rulings and sidewalk delivery robot accidents to demonstrate that physical AI is rapidly entering the exclusion phase. He concludes by presenting Clanker Labs' mission to build the telemetry and inspection standards required to price and safely scale embodied robotics.

## Atmosphere & style The presentation carries a sharp, analytical, and intellectually compelling tone that revitalizes an often overlooked topic into an essential lens for understanding technological revolutions. Acosta delivers his keynote with pragmatic, confident authority, seamlessly connecting dramatic historical catastrophes with modern autonomous hardware mishaps. The pacing feels urgent yet calculated, leaving viewers with a clear and grounded perspective on how underwriter risk models quietly dictate the future of technological deployment.

## Key takeaways - [00:04](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=4s) The landmark Air Canada ruling established that companies cannot treat automated AI agents as separate legal entities to shield themselves from liability. - [01:12](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=72s) Historically, technical safety standards emerge from insurance inspections, beginning with Hartford Steam Boiler turning private checklists into national engineering codes after the 1865 Sultana disaster. - [01:50](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=110s) Catastrophic insurer payouts directly shape global regulatory frameworks, exemplified by marine underwriters mandating 24/7 radio watches and 100% lifeboat capacity under SOLAS after the Titanic sank. - [03:20](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=200s) Property fire insurers founded and funded Underwriters Laboratories (UL) to define hardware safety standards before agreeing to cover risks associated with commercial electrification. - [04:00](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=240s) Commercial nuclear power in the United States was unviable until the federal government legislated capped liability and risk pools, prompting standard homeowner policies to create nuclear exclusions. - [05:47](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=347s) Modern cybersecurity practices like two-factor authentication (2FA) became industry standards primarily because cyber underwriters refused to renew policies without them. - [06:35](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=395s) In response to high-liability incidents involving autonomous robots, insurance carriers are rapidly stripping silent AI coverage and filing explicit machine learning exclusions. - [08:02](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=482s) The evolution of technology insurance follows five distinct stages: silent coverage, panic and exclusions, specialist underwriting, compulsory coverage, and normalization. - [08:35](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=515s) Physical AI and humanoid robotics are currently transitioning from silent coverage into the exclusion phase, creating an urgent need for domain-specific inspection tools and telemetry models. - [09:00](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=540s) Insurance serves as the mechanism by which civilization prices emerging fears, ultimately transforming underwriter checklists into the universal rulebooks for new machines.

## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=4s) The $812 Air Canada Case and the Question of Machine Liability

Insurance is often dismissed as mundane or boring, but understanding its historical evolution is essential for anticipating how liability will shape the future of humanoid robotics. In February 2024, Air Canada found itself in court defending a bizarre situation: its customer service chatbot had invented a bereavement refund policy out of thin air. Rather than admitting a simple software error, Air Canada mounted an audacious defense, arguing that the chatbot was an independent legal entity responsible for its own actions.

The airline lost the case, resulting in a tiny judgment of just $812 Canadian dollars. Despite the modest payout, it became what Darien Acosta describes as "the cheapest landmark case in history." The ruling reopened the foundational dilemma of the automated era: who pays when the machine gets it wrong? Ultimately, insurance steps in to shoulder these losses, but only after underwriters establish rigorous processes to measure, assess, and price the underlying risk.

- **Air Canada Ruling (Feb 2024):** The airline attempted to treat its LLM chatbot as a separate legal entity after it fabricated a discount policy, but the court held the company liable for $812 CAD. - **The Core Precedent:** Automated agents and software cannot serve as liability shields for the corporations deploying them.

---

[01:12](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=72s) Historical Lessons: Steamboat Boilers, the Titanic, and the Birth of Safety Codes

To forecast the trajectory of physical AI, one must look back to the inception of industrial risk management. In 1865, the steamboat *SS Sultana* suffered a catastrophic boiler explosion on the Mississippi River, killing over a thousand passengers in one of the deadliest maritime disasters in American history. In the mid-19th century, unregulated steam boilers exploded with alarming regularity. The following year, in 1866, a group of engineers in Hartford, Connecticut, established the Hartford Steam Boiler Inspection and Insurance Company with an unorthodox mandate: they would only insure a boiler if it passed their strict engineering inspection first.

```mermaid flowchart TD accTitle: The Cycle of Industrial Disasters and Insurance Standardization subgraph Catastrophe ["1. Catastrophic Failure"] A["Industrial Disaster<br/>(e.g., Sultana, Titanic)"] end subgraph Intervention ["2. Underwriter Response"] B["Inspection & Pricing Protocols"] C["Technical Safety Standards"] end subgraph Standardization ["3. Industry Transformation"] D["Insured Standard Becomes<br/>De Facto National Code"] end

A --> B B --> C C --> D ```

Hartford Steam Boiler's private inspection checklist rapidly became the de facto national safety standard across the United States. Almost overnight, catastrophic boiler explosions plummeted, establishing a universal dynamic where "the insured boiler became the safe boiler." Decades later, following the sinking of the *Titanic*, Lloyd’s paid out the massive insurance claims within weeks. To prevent future ruinous payouts, underwriters helped establish the 1914 International Convention for the Safety of Life at Sea (SOLAS), mandating around-the-clock radio watches and sufficient lifeboats for every passenger aboard.

- **1865 *Sultana* Disaster:** Boiler explosion on the Mississippi River caused over 1,000 fatalities, prompting private industry intervention. - **Hartford Steam Boiler (1866):** Created the first pre-insurance inspection checklists, inventing commercial safety compliance. - **Post-*Titanic* SOLAS Standards:** Marine insurers directly shaped international maritime rules, including mandatory 24/7 radio watches and 100% lifeboat capacity.

---

[03:20](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=200s) Electrification, the Nuclear Impasse, and the Invention of Policy Exclusions

Waiting for catastrophes before creating safety standards proved reactive and expensive, prompting insurers to take a proactive approach during the dawn of commercial electricity. To prevent widespread structural fires caused by novel indoor wiring and lighting, fire insurance companies established and funded Underwriters Laboratories (UL). The ubiquitous UL stamp found on modern consumer electronic plugs remains a direct legacy of insurers defining technical hardware safety before agreeing to underwrite property risk.

```mermaid flowchart LR accTitle: Nuclear Liability Impasse and Legislative Solution subgraph Bottleneck ["Uninsurable Risk"] A["Unlimited Reactor<br/>Liability"] --> B["Private Insurers<br/>Refuse Coverage"] B --> C["Utility Construction<br/>Halted"] end subgraph Solution ["Legislative Backstop"] D["Congressional Mandate<br/>(Price-Anderson Act)"] E["Mandatory Insurance Pool<br/>+ Federal Cap"] end subgraph Deployment ["Commercial Scaling"] F["1957: First Commercial<br/>Nuclear Plant (PA)"] G["Nuclear Exclusions Added<br/>to Home Policies"] end

C --> D D --> E E --> F E --> G ```

A different hurdle emerged during the 1950s nuclear boom. While the United Kingdom and Soviet Union rapidly constructed nuclear power plants, the United States stalled because private insurance carriers refused to take on unlimited reactor liability. Without insurance, utilities could not build. Congress intervened by passing legislation to establish a mandatory insurance pool with capped liability and a federal backstop, clearing the path for the first commercial nuclear reactor to open in Pennsylvania in 1957. Simultaneously, insurers introduced the "nuclear exclusion" into standard homeowner policies, cementing the rule that unpriced catastrophic hazards must be carved out into dedicated, specialized lines.

- **Underwriters Laboratories (UL):** Founded and funded by property fire insurers to create standardized safety testing for electrical consumer goods. - **1950s Nuclear Standoff:** Commercial nuclear power in the US was legally blocked until Congress created a mandatory liability pool and federal safety net. - **The Concept of Exclusions:** Standard policies carve out specific perils (e.g., floods, nuclear incidents) to force the creation of specialized underwriting products.

---

[05:47](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=347s) The Evolution of Cyber Insurance and the Rise of Robotic Liability

Policy exclusions systematically give birth to brand-new insurance sectors. For roughly two decades, computer system breaches were silently absorbed under general commercial liability policies. Following massive global cyberattacks in 2017, insurers incurred severe losses and stripped cyber risk out of standard coverage. Standalone cyber insurance emerged, and underwriters quickly imposed technical requirements: companies today implement two-factor authentication (2FA) largely because cyber underwriters refuse to renew policies without it.

```mermaid flowchart TD accTitle: The Five Stages of Insurance Grief subgraph Stage1 ["Stage 1: Silent Coverage"] S1["New machines covered under legacy policies"] end subgraph Stage2 ["Stage 2: Panic & Exclusions"] S2["Losses hit; underwriters write strict exclusions"] end subgraph Stage3 ["Stage 3: Specialist Underwriting"] S3["Specialists inspect hardware & collect telemetry"] end subgraph Stage4 ["Stage 4: Mandated Compliance"] S4["Catastrophe makes coverage compulsory"] end subgraph Stage5 ["Stage 5: Normalization"] S5["Risk is priced into background safety rules"] end

S1 --> S2 S2 --> S3 S3 --> S4 S4 --> S5 ```

A similar transition is beginning for physical AI and robotics. In the fall of 2023, a Starship Technologies autonomous sidewalk delivery robot allegedly reversed into a 73-year-old woman, knocked her down, and struck her again while she was on the ground, causing permanent impairment. Such lawsuits threaten multi-million-dollar "nuclear verdicts" in civil court. In response, silent AI coverage is disappearing rapidly, with 41 insurance companies filing explicit AI exclusions in a single month to disclaim liabilities resulting from machine learning systems.

- **Cyber Insurance Evolution:** General liability covered network attacks until the 2017 cyber catastrophes forced dedicated policies, making 2FA a non-negotiable underwriter mandate. - **Starship Technologies Incident (Fall 2023):** Sidewalk delivery robot allegedly backed over an elderly pedestrian twice, triggering severe personal injury litigation. - **AI Exclusions:** Forty-one major carriers have filed exclusions specifically removing damages caused by machine learning models from standard commercial policies.

---

[08:02](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=482s) The Five Stages of Insurance Grief and Underwriting Physical AI

The historical interaction between emerging technology and risk capital follows a predictable model Acosta terms the "five stages of insurance grief":

1. **Silent Coverage:** The new machine is quietly covered under legacy, general-purpose policies. 2. **Panic and Exclusions:** Unexpected losses hit carriers, leading to panic and blanket policy exclusions. 3. **Specialist Underwriting:** Dedicated underwriters emerge with technical inspection criteria and telemetry tools to price the risk. 4. **Compulsory Coverage:** A high-profile catastrophe or regulatory mandate makes insurance mandatory to operate. 5. **Normalization:** The technology becomes so reliable and standard that society no longer fears it (analogous to modern home boilers).

The physical AI ecosystem currently sits between Stage 1 and Stage 2. A massive influx of competing humanoid robots and autonomous devices is entering warehouses and public spaces without established underwriting frameworks. To fill this gap, Acosta founded Clanker Labs to act as specialist underwriters for the physical AI era. Operating under the motto, *"If it thinks or it moves, we aim to insure it,"* the company's objective is to build the technical inspection criteria and risk models that will govern the coming age of embodied robotics.

- **Civilization's Rulebook:** "Insurance is how civilization prices what it fears, and then the price list becomes the rule book." - **Current Market Phase:** Physical AI is transitioning from silent legacy coverage to widespread exclusions as autonomous robots enter public spaces. - **Clanker Labs:** Founded by Darien Acosta to develop domain-specific underwriting, data collection, and safety checklists for physical AI systems.

## Notable quotes > Who pays when the machine is wrong? It's a question insurance has answered five times in 160 years — Darien Acosta

> the coverage that emerged quietly became the safety standard the technology had to live by. — Darien Acosta

> [00:04](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=4s) the cheapest landmark case in history. — Darien Acosta

> [00:04](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=4s) who pays when the machine gets it wrong? — Darien Acosta

> [01:12](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=72s) the insured boiler became the safe boiler. — Darien Acosta

> [08:02](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=482s) five stages of insurance grief — Darien Acosta

> [08:02](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=482s) If it thinks or it moves, we aim to insure it — Darien Acosta

> [08:02](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=482s) Insurance is how civilization prices what it fears, and then the price list becomes the rule book. — Darien Acosta

## Mentioned Darien Acosta, Clanker Labs, Air Canada, SS Sultana, Hartford Steam Boiler Inspection and Insurance Company, Lloyd's, Titanic, International Convention for the Safety of Life at Sea (SOLAS), Underwriters Laboratories (UL), Price-Anderson Act, Starship Technologies, Ai4

## Video description Darien Acosta - CEO and cofounder, Clanker Labs

Who pays when the machine is wrong? It's a question insurance has answered five times in 160 years — for the steam boiler, the automobile, the aircraft, the power grid, and the networked computer — and each time, the coverage that emerged quietly became the safety standard the technology had to live by. Right now the AI agents running inside your company are losing that coverage as insurers write new exclusions, and the humanoid robots headed for the warehouse floor and the public sidewalk will force the question harder than anything before them. In this talk I'll trace the 160-year pattern that turns every "uninsurable" new machine into an insurable one, and use it to forecast the timeline, the trigger events, and the liability reckoning coming for physical AI. You'll leave able to see the next machine age the way an underwriter does — as a series of predictable, dated events, most of which haven't happened yet. If you build, deploy, insure, or invest in AI, come find out who's really on the hook when the machine gets it wrong

[Source](https://www.youtube.com/watch?v=4ub9YJuNXlY)

中文简报

机器人致伤诉讼威胁巨额赔付 保险公司逐步剔除AI承保责任 随着保险公司为规避巨额事故责任而取消隐性承保，专门的核保清单将成为规范商用机器人部署的事实安全标准。 Darien Acosta 阐述了从19世纪的蒸汽锅炉和电气化到核电及网络风险等重大技术变革，如何在机器发生故障时依赖保险核保人来确立强制性安全规范。他概述了“保险痛苦的五个阶段”，展现了责任如何经历隐性承保、下意识排除，逐步发展到专业核保并最终实现社会常态化。Acosta 将这一框架应用于现代，结合近期关于聊天机器人的法律裁决以及人行道配送机器人事故指出，实体AI正迅速进入排除阶段。他最后介绍了 Clanker Labs 的使命，即建立对具身机器人进行风险定价和安全规模化扩展所需的遥测及检验标准。 加拿大航空（Air Canada）诉讼判决确立了关键法律先例，明确企业无法将自主 AI 智能体视为独立法人实体来逃避自身赔偿责任。 (0:04) 历史上的工业安全标准大多源自保险核保机制：1865 年 Sultana 蒸汽船事故催生了锅炉工程规范，泰坦尼克号沉没则直接促成了海上人命安全公约（SOLAS）的制定。 (1:12) 从火灾保险商出资创立 UL 实验室确立电气硬件标准，到网络安全承保人强制推行双重认证（2FA），核保要求历来会直接转化为行业通用安全规范。 (3:20) 随着人行道配送机器人事故引发高额索赔诉讼，数十家保险公司正密集出台机器学习免责声明，物理 AI 正在经历从传统默示承保走向严格免责与专属核保的周期阶段。 (5:47)

Reading summary

According to Darien Acosta, founder of Clanker Labs, the widespread commercial deployment of humanoid robotics and physical AI will be governed not by software terms of service, but by the insurance sector's capacity to quantify and underwrite machine liability. Acosta argues that historical industrial transitions demonstrate how underwriter safety standards and policy exclusions inevitably become the binding rulebooks for emerging technologies. The Breakdown of Machine Liability Shields (0:04) Corporate attempts to deflect liability onto autonomous software have failed in court. In February 2024, Air Canada contested an $812 Canadian dollar ruling after its customer service chatbot fabricated a bereavement fare policy out of thin air, with the airline arguing the chatbot was a distinct legal entity responsible for its own conduct. Acosta identifies the loss as the cheapest landmark case in history, cementing the legal precedent that automated systems cannot shield operating companies from civil responsibility. This legal exposure intensifies as machine learning models transition from digital text to embodied physical systems. In the fall of 2023, an autonomous sidewalk delivery robot operated by Starship Technologies allegedly struck a 73-year-old woman in reverse and hit her a second time while grounded, causing permanent impairment. Such incidents expose operators to multi-million-dollar nuclear verdicts in civil litigation, reinforcing that corporations will be held strictly accountable when physical automated systems malfunction. Insurers as De Facto Industrial Regulators (1:12) Historically, technical safety codes have emerged from underwriters seeking to curb catastrophic operational losses rather than from government mandates. Following the 1865 SS Sultana disaster—where a steamboat boiler explosion killed over 1,000 passengers—the Hartford Steam Boiler Inspection and Insurance Company was formed in 1866. By requiring pre-coverage engineering inspections, Hartford Steam Boiler's private checklist rapidly evolved into the de facto national safety standard across the United States, proving that the insured equipment became the safe equipment. Underwriters similarly defined marine and electrical safety baselines. After paying out massive claims for the sinking of the Titanic, marine insurers at Lloyd's shaped the 1914 International Convention for the Safety of Life at Sea (SOLAS), mandating continuous 24-hour radio watches and 100 percent lifeboat capacity. To prevent structural fires during the advent of commercial electricity, property fire insurers founded and funded Underwriters Laboratories (UL), instituting hardware inspection stamps before agreeing to write property policies. Carving Out Unpriced Perils Through Exclusions (3:20) When emerging risks present uninsurable catastrophic exposure, standard carriers introduce blanket policy exclusions that necessitate legislative backstops and dedicated underwriting lines. During the 1950s, commercial nuclear power stalled in the United States because private insurers refused unlimited reactor liabilities. Construction resumed only after Congress legislated a mandatory insurance pool with a federal cap, enabling the first commercial plant in Pennsylvania in 1957, while homeowner carriers carved out explicit nuclear exclusions. A matching transition reshaped digital risk. For two decades, computer breaches were silently absorbed under general liability contracts. Severe global cyberattacks in 2017 generated severe losses, prompting underwriters to strip cyber risks out of standard coverage. Standalone cyber insurance emerged as a specialized sector, quickly transforming operational security by making technical requirements, such as two-factor authentication, non-negotiable for policy renewals. The Five Stages of Underwriting Physical AI (8:02) Acosta frames the integration of transformative machinery around five stages of insurance grief: silent coverage under legacy policies, panic and exclusions following unexpected payouts, the emergence of specialist underwriters, compulsory coverage driven by high-profile disasters or regulation, and eventual societal normalization. Physical AI currently sits between stage one and stage two, with legacy coverage collapsing as 41 major insurance carriers filed explicit exclusions in a single month to disclaim damages caused by machine learning models. Clanker Labs was established as a specialist underwriter to navigate this gap, operating under the mandate to insure systems that think or move. Acosta asserts that insurance is civilization's method of pricing fear, which ultimately converts into operational rulebooks. For embodied humanoid robotics, developing proprietary telemetry, hardware inspection standards, and distinct actuarial models will serve as the gateway for commercial operations in shared human environments. Physical AI faces an imminent contraction of legacy liability coverage as carriers enact blanket machine-learning exclusions. Whether specialist underwriters can deploy telemetry and inspection standards quickly enough to avoid a complete commercial bottleneck remains dependent on the sector's ability to model autonomous physical risks before catastrophic accidents trigger compulsory mandates.

精炼阅读摘要

Darien Acosta 认为，具身人工智能与人形机器人的规模化普及不会仅由软件迭代驱动，而将取决于保险业如何衡量、定价与分担物理世界的机器风险；纵观工业历史，承保机制往往通过设立责任除外条款与强制技术安全规范，最终演变为定义新兴技术演进边界的社会守则。 机器法律责任与企业护盾抗辩的破产 (0:04) 在自动化系统逐步介入商业决策与物理世界的过程中，损害赔偿责任的归属成为了首要法律难题。2024年2月，加拿大航空（Air Canada）因其客户服务大型语言模型（LLM）聊天机器人凭空捏造丧亲退款政策而遭到民事起诉。尽管涉案赔偿金额仅为812加元，但该航司在法庭上提出了一项罕见抗辩，主张该聊天机器人是能够独立承担行为后果的法律实体。法院最终驳回了这一主张并判决航司败诉，确立了一项核心判例：企业绝不能将自动化智能体或软件作为隔绝自身法律后果的“责任护盾”。 这一司法界定直接映射到物理人工智能与机器人领域。2023年秋季，Starship Technologies 的自动人行道配送机器人据称在倒车时撞倒了一名73岁女性，并在其倒地后再次实施碾压，导致受害人终身残疾，进而引发了面临数百万美元“天价判决”风险的严重人身伤害诉讼。此类事件表明，不论是虚拟算法还是自主移动实体，其失误代价最终必须由部署企业承担；而当企业无法推卸责任时，损失便会不可避免地转嫁至保险行业，倒逼承保人必须建立严格的风险测量、评估与定价流程。 工业安全范式：从灾难理赔到技术准入标准 (1:12) 现代工业风险管理的发展史表明，技术安全标准的确立往往是被巨额灾难性理赔倒逼出来的。1865年，密西西比河上的“苏丹娜号”（SS Sultana）蒸汽船发生严重锅炉爆炸，造成上千名乘客丧生，成为美国历史上最惨重的海难之一。鉴于19世纪中叶锅炉缺乏监管且爆炸频发，一群工程师于1866年成立了哈特福德蒸汽锅炉检验和保险公司（Hartford Steam Boiler），确立了“唯有通过严格工程检验方可承保”的准入原则。该公司的私人检验清单迅速成为全美事实上的国家工业安全规范，使锅炉爆炸事件急剧减少，进而形成了“通过投保检验即意味着安全”的行业常态。 这种承保机制塑造行业标准的逻辑在此后的重大工业演进中一再重现。1912年“泰坦尼克号”沉没后，英国劳合社（Lloyd's）在数周内全额赔付了巨额理赔金；为遏制灾难性财务亏损重演，海上承保人深度参与并推动了1914年《国际海上人命安全公约》（SOLAS）的出台，强制推行24小时全天候无线电值守与100%救生艇定额配置。此后在商业电气化初期，为防止室内布线与照明引发普遍性火灾，财产火险公司更出资设立了保险商实验室（UL），使消费电子产品上的 UL 安全认证标志至今仍是承保前置的核心硬件基准。 风险除外与国家介入：从核能到网络安全 (3:20) 当新技术面临可能导致承保机构破产且无法精算定价的极端风险时，保险业便会采取“责任除外”（Exclusion）策略进行切割，进而迫使国家立法介入或催生独立的专门险种。20世纪50年代美国核能商业化一度陷入停滞，根本原因在于私营保险商拒绝为核反应堆承担无限赔偿责任，导致电力公用事业机构无法启动建设。直至美国国会通过专门法案确立强制保险池、设定赔偿限额并提供联邦兜底后，全美首座商业核电站才于1957年在宾夕法尼亚州获准建成；与此同时，承保人正式在标准房屋所有人体险保单中引入“核风险除外条款”，奠定了将灾难性风险剥离为专项险种的运作惯例。 网络安全保险的演进也重现了这一模式。在长达二十年的时间里，计算机系统入侵风险均以“沉默覆盖”（Silent Coverage）的形式被传统商业综合责任险顺带消化。直到2017年多起全球性重大网络攻击造成巨额亏损后，承保人才迅速从基础保单中剔除网络安全风险，催生了独立的网络保险市场。这种剥离直接构筑了新的技术安全底线：当前多数网络安全承保人已将部署双重身份验证（2FA）定为不可妥协的承保与续保前提。历史经验证明，除外条款并不是技术发展的终点，而是倒逼专业风控规范成型的分水岭。 具身智能悲伤五阶段与专业承保的演进 (8:02) Darien Acosta 提出了新兴技术与风险资本互动的“保险悲伤五阶段”演进框架：第一阶段为“沉默覆盖”，即新技术在初期被传统通用保单无意识接纳；第二阶段为“恐慌与除外”，突发性亏损促使承保人出台全面的免责除外声明；第三阶段为“专业承保”，具备技术检验与遥测数据采集能力的专业机构入场为风险定价；第四阶段为“强制覆盖”，恶性灾难或监管法案使投保成为运营的强制前提；第五阶段为“常态化”，技术安全性完全被吸纳为社会底层的平淡常识。 当前，物理 AI 与人形机器人产业正处于从“沉默覆盖”向“恐慌与除外”滑落的过渡阶段。随着大量自主设备进入仓库与公共空间，通用保险公司单月内便有41家集中提交了针对机器学习系统的明确除外条款，拒绝承担由算法失误导致的损害。为填补这一传统保险退潮后的承保真空，Acosta 创立了 Clanker Labs，秉持“凡有思考或运动特征者皆当受保”的目标，致力于针对具身机器人开发专属的技术检验清单、遥测数据模型与专业承保方案。这印证了其核心洞见：保险是人类文明为所恐惧之物定价的手段，而最终这份定价标准往往会演变为全行业的运作规则。 物理 AI 的发展正在重演从工业锅炉到网络系统曾经历的承保周期，但机器人自主移动与黑箱决策的结合带来了前所未有的责任复杂性。当前通用保险市场正迅速关闭对 AI 风险的沉默覆盖，然而在缺乏类似核工业联邦立法限额与兜底机制的背景下，初创专业承保机构仅凭遥测工具与工程清单能否有效抵御民事诉讼中的巨额赔偿诉求，并在行业发生重大伤亡灾难前确立统一的安全合规标准，仍是悬而未决的严峻考验。

Detailed notes — zh

## Background 本内容由频道 Ai4 于 2026 年 8 月 15 日在 YouTube 平台发布。视频记录了 Clanker Labs 首席执行官兼联合创始人 Darien Acosta 的主题演讲，探讨了过去 160 年来保险业如何将蒸汽锅炉、汽车、网络等高风险技术逐步转化为安全标准，并以此预测具身智能与人形机器人即将面临的责任归属与承保变革。 ## TL;DR 演讲从加拿大航空 Chatbot 诉讼与配送机器人致残案切入，指出当机器犯错时责任最终需由保险机制吸纳。回顾蒸汽锅炉检验、泰坦尼克海难后的 SOLAS 公约、核能责任法案以及网络安全险的演进，保险业始终遵循从“隐性承保”、遭遇亏损设立“除外责任”，到专业核保人介入建立行业安全标准的固定周期。当前已有数十家险企出台 AI 除外条款，具身智能正步入责任剥离的关键转折点，Clanker Labs 正试图通过建立专业核保模型填补物理 AI 领域的风险定价空白。 ## Atmosphere & style 整场演讲冷静而富有洞察力，主讲人以核保人严谨理性的视角重新审视前沿科技的落地路径。叙事风格兼具历史厚重感与对当下 AI 产业的警示感，语言凝练且直击痛点，将看似枯燥的保险条款转化为决定人形机器人商业命运的底层逻辑，给听众带来一种看透技术浪潮演变规律的笃定与启发。 ## Key takeaways - [00:04](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=4s) 加拿大航空客服 Chatbot 败诉案驳回了“AI 是对其行为负责的独立法律实体”的主张，再次引发工业社会中“机器犯错谁来买单”的根本追问。 - [00:04](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=4s) 技术造成的损失最终通常由保险机制吸纳，但其成立的前提是市场必须先形成一套完整的风险评估与定价机制。 - [01:12](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=72s) 1865 年苏丹娜号蒸汽轮船爆炸催生了哈特福德公司“先检验、后承保”模式，使保险核保清单逐步演变为事实上的国家工业安全规范。 - [01:12](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=72s) 从泰坦尼克号推动 SOLAS 海事公约到火灾保险商创立 UL 认证，保险机构的核保标准历来都会直接固化为全行业的出厂与运营安全基线。 - [04:02](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=242s) 1957 年美国通过设定赔偿限额与联邦兜底机制化解了核反应堆无限责任的僵局，使不可保的高危技术得以商业化落地。 - [04:02](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=242s) 保险业发明的“除外责任”（Exclusion）条款通过将无法承担的巨灾风险从通用保单剥离，倒逼并催生了独立的专业险种与细分市场。 - [04:02](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=242s) 网络安全险经历了 20 年的“隐性承保”才在巨额损失后独立成险，并反向将双重身份验证（2FA）等风控手段确立为全行业的技术准则。 - [06:32](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=392s) 具身智能在公共空间落地引发严重的人身侵权风险，如配送机器人撞人致残案极易进入陪审团裁决并引发数百万美元的“核级别判决”。 - [06:32](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=392s) 传统保险公司正加速关闭风险敞口，已有 41 家险企启动“AI 除外责任”，机器学习造成的物理与财产损失正迅速丧失通用保单覆盖。 - [08:02](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=482s) 新技术在保险演进中遵循固定的“悲伤五阶段”：从隐性承保、损失爆发设立除外，逐步演化至专业核保、立法强制并最终成为社会安全常态。 - [08:02](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=482s) 保险是人类文明为其恐惧定价的方式，而由承保人制定的风险价格清单最终会转化为新技术产业必须遵循的行为规则守则。 - [08:02](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=482s) 具身智能正处于从隐性承保向除外责任过渡的关键窗口，Clanker Labs 等机构正在构建物理 AI 领域的专业核保与风险定价模型。 ## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=4s) 谁该为机器的过错买单？从加拿大航空 Chatbot 诉讼谈起 保险在大多数人看来可能枯燥乏味，但要理解具身智能与人形机器人（Humanoid Robotics）的未来，唯一的钥匙就是回顾保险业的历史。2024 年 2 月，加拿大航空（Air Canada）在一场法庭诉讼中辩称，其客服 Chatbot 凭空捏造了一项退票政策，航空公司主张该 Chatbot 是一个“对自身行为负责的独立法律实体”（a separate legal entity responsible for its own actions）。这场辩词堪称大胆，但加拿大航空最终败诉，被判赔偿 812 加元。虽然这可能是史上赔付金额最小的标志性案件，但它重新引出了现代工业社会中最古老的问题：“当机器犯错时，到底谁来买单？”（Who pays when the machine is wrong?） 最终的答案通常是“由保险买单”，但前提是必须存在一套评估风险与定价风险的完整机制。要洞察未来的机器人时代，就必须先理解这套定价机制在历史上是如何形成的。 - **加拿大航空退票案**：因客服 Chatbot 虚构退票政策引发诉讼，法院驳回了“AI 为独立法律主体”的主张，判赔 812 加元。 - **机器责任的核心逻辑**：技术造成的损失最终由保险吸纳，但必须经历风险量化与定价流程的建立。 --- ### [01:12](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=72s) 蒸汽锅炉、泰坦尼克与 UL：保险如何化身工业安全标准 回顾现代保险的起点，1865 年密西西比河上的“苏丹娜号”（Sultana）蒸汽轮船发生锅炉爆炸，造成 1000 多人死亡，成为美国历史上最严重的海难之一。在 19 世纪，蒸汽锅炉缺乏监管，爆炸事故几乎每天都在发生。次年（1866 年），康涅狄格州哈特福德的一群工程师创立了哈特福德蒸汽锅炉检验和保险公司（Hartford Steam Boiler），并提出了一种全新的商业模式：“我们先检查你的锅炉，只有通过检查，我们才予以承保。”这套检验清单随后演变为事实上的美国国家安全规范，锅炉爆炸率随即暴跌，形成了“被承保的锅炉就是安全的锅炉”的行业认知。 灾难往往倒逼监管与标准的诞生。1912 年泰坦尼克号沉没后，英国劳合社（Lloyd's）在数周内完成了理赔，并在两年后联合各方促成了 1914 年《国际海上人命安全公约》（SOLAS）的签署，强制要求配备满足所有乘客人额的救生艇以及全天候无线电值守。为了不再被动等待灾难发生，保险业在电气化时代转向了主动预防：由火灾保险商出资成立的保险商实验室（Underwriters Laboratories，即 UL），其制定的标准至今仍以 “UL” 标识印在几乎所有家用电器的插头上。 - **1865 年苏丹娜号轮船爆炸**：千余人丧生，催生了 1866 年哈特福德蒸汽锅炉公司的“先检验、后承保”机制。 - **泰坦尼克号与 SOLAS 公约**：推动了船舶全员救生艇配额与 24 小时无线电监控的全球海事安全标准。 - **UL 安全认证**：由火灾保险机构资助创立，将保险核保标准直接固化为消费电子产品的出厂安全规范。 --- ### [04:02](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=242s) 核能法案、保单除外责任与网络安全保险的演进 当一个行业因内在的高风险性而无法依靠纯市场化机制扩张时，保险往往成为瓶颈。20 世纪 50 年代核能发展初期，英国和苏联迅速推进核电站建设，而美国却因责任法限制停滞不前——没有任何私营保险公司愿意承担反应堆的无限责任。为此，美国国会立法建立了强制保险池、赔偿上限以及联邦托底机制，最终促成 1957 年宾夕法尼亚州首座商用核反应堆的投运。 与此同时，保险业在 1957 年发明了“除外责任”（Exclusion）条款，在所有屋主财产险保单中加入了“核事故除外责任”——若社区发生核熔毁导致房屋全损，普通房屋保险概不负责。“除外责任”将原本被认为应受保障的风险剥离出去，进而催生出全新的独立险种（如洪水险、核能险）。类似的演变在网络安全险（Cyber Insurance）中同样发生过：在长达 20 年的时间里，计算机黑客攻击一直被混在通用商业综合责任险（General Liability）中“隐性承保”（Silent Coverage）；直到 2017 年全球性大规模黑客攻击爆发造成巨额损失，保险业才全面引入网络安全除外条款，强制企业购买独立网络险。如今企业普遍推行的双重身份验证（2FA），正是网络安全核保人在保单续保时强制推行的风控准则。 ```mermaid flowchart TD accTitle: Insurance Exclusions and Specialized Lines Evolution accDescription: Flowchart illustrating how uninsurable catastrophic risks lead to policy exclusions and eventually create specialized insurance lines and safety standards. subgraph Phase1["隐性承保与风险暴露"] A["新技术出现<br/>（核能 / 早期计算机）"] --> B["混入传统保单<br/>（普通房屋险 / 综合责任险）"] B --> C["系统性灾难/重大损失爆发<br/>（2017 全球网络攻击）"] end subgraph Phase2["责任剥离与专用险种"] C --> D["保单写入除外责任<br/>（Exclusion 条款）"] D --> E["催生独立专用险种<br/>（独立网络安全险 / 独立核能险）"] end subgraph Phase3["标准固化"] E --> F["核保人强制设立安全基线<br/>（如强制推行 2FA 认证）"] end ``` - **核能责任法案（1957 年）**：通过设定赔偿限额与联邦兜底，打破核电产业因无限责任无法承保的死局。 - **除外责任（Exclusion）的作用**：剥离系统性巨灾风险，倒逼市场建立独立的风险定价通道与新险种。 - **网络安全险的历史重演**：从传统责任险的混杂覆盖，演进到 2017 年后的独立险种，并最终反向定义了企业 IT 的安全技术标准（如 2FA）。 --- ### [06:32](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=392s) 具身智能的隐性承保与正在逼近的人形机器人诉讼浪潮 在具身智能（Physical AI）与人形机器人领域，当前行业整体仍处于“隐性承保”（Silent Coverage）阶段，但涉及机器人的侵权诉讼已经出现且呈加速趋势，保险除外责任的到来已不可逆转。2023 年秋季，Starship Technologies 的配送机器人在人行道上倒车时撞倒了一名 73 岁女性，并在其倒地后再次发生碰撞碾压，导致该受害者永久性伤残。这类案件若进入陪审团裁决，极易触发包含高额精神损害与惩罚性赔偿的“核级别判决”（Nuclear Verdict），索赔金额可达数百万美元。 面对这种不可预测的物理伤害与巨额索赔，传统保险公司的必然反应就是出台除外条款。截至演讲当月的前一个月，已有 41 家保险公司向监管部门提交并启用了“AI 除外责任”（AI Exclusion），明确将任何由机器学习系统造成的损失从标准商业保单中剔除。AI 与机器人的“隐性承保时代”正在迅速终结。 - **Starship Technologies 人行道机器人侵权案**：2023 年秋发生的配送机器人碰撞致残诉讼，凸显物理 AI 设备在公共空间的致害风险。 - **核级别判决（Nuclear Verdict）风险**：惩罚性赔偿与人身伤害赔偿叠加，促使传统承保机构快速收紧风险敞口。 - **41 家险企启动 AI 除外条款**：主流保险公司正加速将机器学习系统的致害责任彻底剥离出通用保单。 --- ### [08:02](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=482s) “保险悲伤五阶段”与 Clanker Labs 的具身 AI 承保蓝图 根据 160 年来的工业技术演进史，新技术与保险之间存在着固定的“保险悲伤五阶段”（Five Stages of Insurance Grief）。正如达里恩·阿科斯塔（Darien Acosta）所总结：“保险是人类文明为其恐惧定价的方式，而这张价格清单最终会演变成行业的规则守则”（Insurance is how civilization prices what it fears and then the price list becomes the rule book）。 ```mermaid flowchart LR accTitle: The Five Stages of Insurance Grief accDescription: Five chronological stages showing how emerging dangerous technologies become safely insured and standardized. S1["第一阶段<br/>旧保单隐性承保<br/>(Silent Coverage)"] --> S2["第二阶段<br/>损失爆发险企亏损<br/>设立除外责任 (Exclusion)"] S2 --> S3["第三阶段<br/>专业核保人入场<br/>制定检验标准与数据收集"] S3 --> S4["第四阶段<br/>发生重大灾难<br/>推动专用保险强制化"] S4 --> S5["第五阶段<br/>技术高度安全化<br/>融入日常社会基础设施"] ``` 当前具身智能与人形机器人正处于第一阶段向第二阶段过渡的临界点：大量人形机器人和各类自动化移动设备即将涌入仓储、工厂与公共人行道，但行业至今缺乏标准化的核保与风险评估模型。为了填补这一空白，Darien 创立了专注于具身智能专业承保的 Clanker Labs，致力于成为物理 AI 领域的专业核保机构。其核心口号是：“只要它能思考，或者它会移动，我们就要承保它”（If it thinks or it moves, we aim to insure it）。 | 阶段 | 核心特征 | 历史案例映射 | 具身智能当前进度 | |:--- |:--- |:--- |:--- | | **第一阶段** | 新机器被混在旧保单中隐性承保 | 早期锅炉爆炸、2017 年前的计算机黑客攻击 | 人形机器人处于本阶段尾声 | | **第二阶段** | 损失集中爆发，保险公司恐慌并全面引入除外条款 | 1957 年核能除外、41 家险企设立 AI 除外责任 | 正加速进入（AI 除外条款落地） | | **第三阶段** | 专用险种诞生，专业核保人携带检验清单和数据模型介入 | 1866 年哈特福德锅炉检验、现代网络安全核保 | Clanker Labs 正在布局介入 | | **第四阶段** | 标志性重大事故迫使该险种与安全标准成为法律强制要求 | 1912 年泰坦尼克号公约、核能责任法案 | 尚未发生 | | **第五阶段** | 技术高度安全化，公众对其安全性习以为常 | 现代建筑内的安全锅炉、带 UL 标的插头 | 远期终局目标 | ## Notable quotes > [00:04](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=4s) Who pays when the machine is wrong? — Darien Acosta > [00:04](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=4s) a separate legal entity responsible for its own actions — Air Canada (in court filing) > [01:12](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=72s) 我们先检查你的锅炉，只有通过检查，我们才予以承保。 — Hartford Steam Boiler > [01:12](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=72s) 被承保的锅炉就是安全的锅炉 — Darien Acosta > [08:02](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=482s) Insurance is how civilization prices what it fears and then the price list becomes the rule book — Darien Acosta > [08:02](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=482s) If it thinks or it moves, we aim to insure it — Clanker Labs > [08:02](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=482s) 只要它能思考，或者它会移动，我们就要承保它 — Clanker Labs > [08:02](https://www.youtube.com/watch?v=4ub9YJuNXlY&t=482s) 保险是人类文明为其恐惧定价的方式，而这张价格清单最终会演变成行业的规则守则 — Darien Acosta ## Mentioned Darien Acosta, Clanker Labs, Air Canada (加拿大航空), 苏丹娜号 (Sultana), Hartford Steam Boiler (哈特福德蒸汽锅炉检验和保险公司), 泰坦尼克号 (Titanic), Lloyd's (劳合社), SOLAS (国际海上人命安全公约), Underwriters Laboratories (UL / 保险商实验室), Starship Technologies ## Video description Darien Acosta - CEO and cofounder, Clanker Labs Who pays when the machine is wrong? It's a question insurance has answered five times in 160 years — for the steam boiler, the automobile, the aircraft, the power grid, and the networked computer — and each time, the coverage that emerged quietly became the safety standard the technology had to live by. Right now the AI agents running inside your company are losing that coverage as insurers write new exclusions, and the humanoid robots headed for the warehouse floor and the public sidewalk will force the question harder than anything before them. In this talk I'll trace the 160-year pattern that turns every "uninsurable" new machine into an insurable one, and use it to forecast the timeline, the trigger events, and the liability reckoning coming for physical AI. You'll leave able to see the next machine age the way an underwriter does — as a series of predictable, dated events, most of which haven't happened yet. If you build, deploy, insure, or invest in AI, come find out who's really on the hook when the machine gets it wrong [Source](https://www.youtube.com/watch?v=4ub9YJuNXlY)

Source quotes

[0:04] Darien Acosta: Who pays when the machine is wrong? [0:04] Air Canada (in court filing): a separate legal entity responsible for its own actions [1:12] Hartford Steam Boiler: 我们先检查你的锅炉，只有通过检查，我们才予以承保。 [1:12] Darien Acosta: 被承保的锅炉就是安全的锅炉 [8:02] Darien Acosta: Insurance is how civilization prices what it fears and then the price list becomes the rule book [8:02] Clanker Labs: If it thinks or it moves, we aim to insure it [8:02] Clanker Labs: 只要它能思考，或者它会移动，我们就要承保它 [8:02] Darien Acosta: 保险是人类文明为其恐惧定价的方式，而这张价格清单最终会演变成行业的规则守则

Source quotes

Darien Acosta: Who pays when the machine is wrong? It's a question insurance has answered five times in 160 years Darien Acosta: the coverage that emerged quietly became the safety standard the technology had to live by. [0:04] Darien Acosta: the cheapest landmark case in history. [0:04] Darien Acosta: who pays when the machine gets it wrong? [1:12] Darien Acosta: the insured boiler became the safe boiler. [8:02] Darien Acosta: five stages of insurance grief [8:02] Darien Acosta: If it thinks or it moves, we aim to insure it [8:02] Darien Acosta: Insurance is how civilization prices what it fears, and then the price list becomes the rule book.

Original transcript

[0:04] insurance is generally a bit boring but I I think  the way to understand it better and how it would [0:11] affect uh the future in terms of humanoid robotics  is really to understand uh the history. Um, [0:20] so in February 2024, uh, Air Canada stood in court  and made an argument that its customer service [0:27] chatbot had invented a refund policy out of thin  air. Air Canada's defense was that the chatbot was [0:35] a separate legal entity responsible for its own  actions. Audacious, they lost. The damages though [0:44] very small, $812 uh, Canadian dollars. Uh this  was the cheapest landmark case in history. But [0:51] it reopened the oldest question in the modern age.  Who pays when the machine is wrong? The answer is [0:58] usually eventually insurance pays but only when  there's a process to assess and price risk. [1:07] In general to understand the future,  you need to uh understand the past. [1:12] So let's go all the way back to very beginnings of  insurance. 1865 uh a steamboat called the Sultana [1:22] uh exploded there uh its boilers uh  exploded on uh on the Mississippi. [1:29] Uh one of the worst maritime disasters in  American history. Over a thousand dead. [1:34] We all hear about the Titanic, but we  don't necessarily hear about this. Uh [1:40] but it had a huge impact on the  insurance industry. The very next year [1:46] um a group of engineers in Hartford, Connecticut  founded a company with a strange business model. [1:51] They said, "We're going to inspect your boiler,  and only if it passes inspection do we insure it." [1:58] uh boilers at during that t during those times  were kind of unregulated uh blowing up every [2:04] day and you know it was the 1800s no one really  cared much um but Harford steam boilers inspection [2:13] uh checklist became the de facto national  uh code uh and once they implemented uh [2:20] the this approach explosions collapsed suddenly  the insured boiler became the safe boil boiler [2:28] Then we go to uh another disaster uh  Titanic. Um we all know the this the story. [2:37] Um the insurance company Lloyds paid  the insurance claims within a few weeks. [2:42] Uh but uh in order to prevent such a catastrophe  from happening again uh 2 years later [2:49] uh there was a convention signed safety at at life  at sea uh with inputs from uh Lloyd's register. [2:58] This set the standards for lifeboats on every  uh for uh for every single passenger uh radio [3:05] uh watch around the clock. In in general, this  was to make uh cruise ships and maritime uh travel [3:13] uh safer. Uh but uh disasters leading to  regulation and safety is generally not ideal. [3:20] So um insurers decided to try to be proactive  with the uh dawn of uh electricity um [3:30] basically wiring lighting in everyone's houses um  etc etc. Um the UL that is stamped on literally [3:41] every plug that exists in the country stands for  underwriters laboratory. This is a safety mark [3:48] on every device that you own. Uh and it's named  after uh after the insurance industry and it was [3:54] funded by fire insurers to prevent house fires.  Um and this this standard exists to this day. [4:02] Um and there are cases where sometimes um  an industry wants to scale but you can't [4:11] it's impossible to scale it because it's  inherently uh dangerous. We saw this with uh [4:16] the dawn of the nuclear age in 19 in the 1950s.  Um the UK was building power plants, Russia was [4:25] building power plants but the US was actually  falling behind because of the the liability law. [4:31] No private insurance insurance carrier would  take unlimited reactor liability. So no utility [4:37] could be built. Uh this warranted an act from  Congress to set up a mandatory insurance pole, [4:45] a maximum cap in a federal backs stop.  And then later in 1957, the first uh [4:51] commercial nuclear reactor opens in Pennsylvania.  That's a that's a picture of what it looked like. [4:57] Um but there was also a side effect of this.  Uh the same year the uh insurance industry [5:04] came up with this idea called an exclusion  uh and wrote nuclear exclusion in everyone's [5:11] uh home insurance policy which is basically to say  if there's a nuclear meltdown in your neighborhood [5:16] and you and your house is a total  loss we don't cover that. Um, [5:22] exclusion is a very interesting concept  because it's basically taking what you [5:29] believe should be covered um and then excluding  it from the policy and forcing you to get uh [5:35] separate insurance. This is similar to like uh  you can't get uh can't be covered for flooding [5:41] under your house. You need separate flood  insurance. Uh but exclusions birth the uh [5:47] new types of insurance lines. For example, we saw  this with cyber insurance uh for 20 years. Uh um [5:56] computer systems could be could be hacked and that  would be covered under standard general liability. [6:01] Uh but then uh in 2017 there was a gigantic hack  that affected everyone and then uh suddenly uh [6:10] cyber insurance needed to be a separate policy.  Um, so today if your company uh runs two-factor uh [6:18] authentification to FA, it's probably because  your cyber underwriter refuse to renew uh your [6:25] company's policy without adhering to to that. Um  so I think you're starting to see the pattern. [6:32] Um with with regard to physical AI robotics,  um we're generally in an age of like [6:40] silent coverage. Uh but the robotic lawsuits  are already here uh and accelerating uh and [6:49] I'm predicting that exclusions are coming. Um so  fall in 2023 uh Starship uh technology sidewalk [6:57] robot uh allegedly reverses into a 73 year old  woman, knocking her down and then striking her [7:03] again while she's on the ground. alleged uh uh  injuries uh leading to permanent impairment and um [7:12] the potential claims o of of this sort of case  uh if it reaches what's called a nuclear verdict [7:18] which is like um pain and suffering uh punitive  damages could cost uh the company millions. So [7:27] um the predicted insurance reaction to this sort  of outcome is going to be exclusion. And actually [7:35] as of uh last month 41 insurance companies  have already adopted to file what they call AI [7:42] exclusion which is basically any um any loss that  is attributed to a machine learning system. Um, so [7:55] we're still kind of in a period of what we call  silent AI coverage, but but this is going away. So [8:02] in general, I see this pattern over and over  again. I I call it the five stages of insurance [8:10] grief. Stage one, uh the new m new machine is  uh silently covered in old policies. Stage two, [8:18] losses hit. Uh insurance lose money,  they panic and then adopt exclusions. [8:24] Uh stage three, uh special uh separate line  require specialist underwriters who show up with [8:30] uh inspection criteria and start collecting  data to manage risk. Uh and then stage four [8:37] maybe a gigantic catastrophe happens that makes  that coverage uh compulsory. And then stage [8:44] five is basically this becomes a safe uh safe  thing that no one even uh cares about. Uh such [8:50] as like we don't worry about our boilers and our  house blowing up. Insurance is how civilization [8:56] prices what it fears and then the price list  becomes the rule book. So uh where are we now? [9:02] Um I think we're in between stage one and stage  two. uh physical AI market is about to be flooded [9:08] with competing uh humanoid devices and other  types of robotics. Uh and there's still no [9:14] standard underwriting approach. Um so what what  do I do? Um I set up a company which I'm calling [9:22] uh Clanker Labs and we basically aim to be  special specialist underwriters for physical AI. [9:29] Um and yeah that our our slogan is if it  thinks or it moves, we aim to ensure it. Um, [9:35] and if you're interested in in any of this,  like find me on LinkedIn, come talk to me, [9:40] and yeah, uh, I'll make it, you know, we'll have  a a a fun conversation at least. Okay. Thank you.