IOC Deploys AI to Block Athlete Abuse, Cut Judging Speed Under 450ms

Inside the Olympic AI Agenda: Blueprint for Transformation with International Olympic Committee — Ai4

Ai4

The International Olympic Committee is standardizing AI governance to scale production systems across officiating, mobile scouting, and cyber-abuse protection before tech divides widen across global sports.

AI-assisted summary of Inside the Olympic AI Agenda: Blueprint for Transformation with International Olympic Committee — original source. Check the original for full context.

Source published:

Detailed notes

## Background Published on YouTube by the Ai4 channel on August 21, 2026, this presentation features Alejandro Merino-Madrid, Head of AI at the International Olympic Committee (IOC). The talk outlines how AI serves as a system-level transformation engine across the Olympic Movement, walking through the Olympic AI Agenda's five strategic focus areas, real-world case studies like digital twins, and frameworks for balancing bold innovation with strict governance.

## TL;DR Alejandro Merino-Madrid outlines the International Olympic Committee's strategic blueprint for artificial intelligence, structured around five core focus areas designed to align technical innovation with foundational Olympic values. He showcases tangible deployments, including the AI Sentinel system that protected over 15,000 athletes from online harassment at Paris 2024, tablet-based computer vision talent scouting deployed in Senegal, and sub-450ms digital twins powering live judging and broadcast analytics. To govern this ecosystem responsibly, the IOC collaborated with Deloitte to establish a practical Trustworthy AI Framework and evaluation checklist for sporting federations. Ultimately, Merino-Madrid shares hard-won lessons on data readiness, proactive governance, and role-based access, urging sports bodies to move beyond endless proofs of concept to build production-grade systems with strong ethical guardrails.

## Atmosphere & style The presentation carries an authoritative, executive, and purposeful tone, seamlessly blending organizational strategy with concrete operational candor. Merino-Madrid grounds high-level corporate frameworks with poignant, human-centric stories—such as protecting vulnerable flagbearers from coordinated digital hate and validating real-time digital twin latency during high-stakes figure skating falls. The delivery feels pragmatic and clear-eyed rather than speculative hype, giving the impression of an authentic masterclass in navigating large-scale enterprise transformation within a complex global institution.

## Key takeaways - [00:04](https://www.youtube.com/watch?v=BFiECev5w7s&t=4s) The IOC structures its AI strategy around Simon Sinek's 'Why, How, What' Golden Circle framework to proactively guide unavoidable technological shifts while safeguarding core Olympic values. - [00:04](https://www.youtube.com/watch?v=BFiECev5w7s&t=4s) The Olympic AI Agenda establishes five strategic focus areas: supporting athletes, ensuring equitable access, optimizing operations, elevating fan experiences, and driving internal efficiency. - [03:04](https://www.youtube.com/watch?v=BFiECev5w7s&t=184s) The AI Sentinel system proactively safeguards athletes' mental well-being by monitoring billions of social posts and removing targeted hate speech before athletes or their families encounter it. - [03:04](https://www.youtube.com/watch?v=BFiECev5w7s&t=184s) Targeted online toxicity during global sporting events is rising rapidly, with Milano Cortina 2026 seeing over a 40% surge in abuse despite lower total engagement compared to Summer Games. - [06:31](https://www.youtube.com/watch?v=BFiECev5w7s&t=391s) Portable computer vision tools democratize talent identification by evaluating youth in remote regions like Senegal, discovering high-potential athletes isolated from traditional scouting networks. - [08:31](https://www.youtube.com/watch?v=BFiECev5w7s&t=511s) Sub-second digital twin technology processes live biomechanical data in under 450 milliseconds to deliver real-time technical feedback, broadcast overlays, and officiating precision. - [11:20](https://www.youtube.com/watch?v=BFiECev5w7s&t=680s) The IOC and Deloitte developed a Trustworthy AI Framework featuring practical governance checklists to help smaller sports organizations evaluate tech vendors safely without legal or compliance risks. - [13:35](https://www.youtube.com/watch?v=BFiECev5w7s&t=815s) Proactive data preparation and cleaning must precede model implementation to prevent legacy and unstructured archives from derailing AI deployments. - [13:35](https://www.youtube.com/watch?v=BFiECev5w7s&t=815s) Deploying enterprise LLM agents requires strict Role-Based Access Control (RBAC) to eliminate internal data leakage and manage complex cross-jurisdictional compliance. - [16:17](https://www.youtube.com/watch?v=BFiECev5w7s&t=977s) Sustainable AI transformation requires organizations to move past disconnected 'proof of concepts' and focus on shipping integrated, production-ready solutions backed by strong guardrails.

## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=BFiECev5w7s&t=4s) Why, How, and What: The Golden Circle for Olympic AI

Drawing inspiration from Simon Sinek's "Start with Why" Golden Circle framework, the International Olympic Committee (IOC) structures its AI transformation strategy around three fundamental questions: why the Olympic Movement requires an AI agenda, how it plans to advance toward that ambition, and what tangible initiatives are being deployed. The underlying reality driving this effort is that artificial intelligence will inevitably transform the sports industry "whether we like it or not and whether we are ready or not." Rather than passively waiting and hoping the technology yields positive outcomes, the IOC has chosen to actively steer the transformation to ensure it aligns with foundational Olympic values: fairness, excellence, and solidarity.

To guide this ambition, the IOC established five distinct focus areas that act as a strategic filter and North Star for funding and development:

```mermaid flowchart TD accTitle: Olympic AI Agenda Five Focus Areas subgraph Core["Olympic AI Agenda Focus Areas"] F1["1. Support Athletes & Competition"] F2["2. Ensure Equitable Access"] F3["3. Optimize Games Operations"] F4["4. Elevate Fan Experience"] F5["5. Drive Internal Efficiency"] end F1 --> Core F2 --> Core F3 --> Core F4 --> Core F5 --> Core ```

- **Supporting Athletes, Competition, and Sport:** Prioritizing athlete well-being, performance, and integrity above all else. - **Ensuring Equitable Access:** Preventing technological division across sports organizations by making AI benefits universally available. - **Optimizing Games Operations:** Delivering Olympic Games faster, more sustainably, and more efficiently. - **Transforming Fan Experiences:** Engaging existing audiences and reaching newer generations through interactive broadcast and venue tools. - **Enhancing Internal Efficiency:** Looking inward to optimize internal operations across the IOC and its partner ecosystem.

---

[03:04](https://www.youtube.com/watch?v=BFiECev5w7s&t=184s) Focus Area 1: Safeguarding Athletes from Online Abuse

Athletes dedicate their lives to reach the Olympic Games, yet they increasingly face targeted digital toxicity, hate speech, and harassment based on race, religion, ethnicity, or political beliefs. To address this crisis, the IOC developed an AI Sentinel system designed to scan and filter billions of online posts across social platforms, automatically identifying and reporting abusive content before athletes or their families ever see it.

During the Paris 2024 Olympic Games, the AI Sentinel monitored billions of posts and comments, identifying 2.4 million messages that directly referenced athletes, officials, or teams. While the majority were standard fan commentary, over 10,000 hate speech posts were flagged and submitted for immediate removal. The protection service shielded 15,000 athletes across both Summer and Winter Games, providing critical psychosocial relief to 353 athletes who were actively targeted by coordinated cyber-abuse campaigns.

The necessity of the tool became even clearer during the Milano Cortina 2026 Winter Games: despite winter events seeing roughly 20% lower overall engagement compared to summer editions, instances of online abuse and hate speech surged by more than 40%. Maxine Egner, the flagbearer for Botswana at Paris 2024, recounted receiving a barrage of hateful messages and targeted images, emphasizing that the IOC's cyber-abuse protection allowed athletes to step away from online toxicity, safeguard their mental well-being, and "actually focus on who you are" and your performance on the field of play.

---

[06:31](https://www.youtube.com/watch?v=BFiECev5w7s&t=391s) Focus Area 2: Democratizing Talent Scouting and AI Access

Left unmanaged, advanced AI risks widening the competitive gap between wealthy organizations with high AI literacy and under-resourced sporting bodies. To counter this divide, the IOC is applying computer vision and athletic assessment models to democratize talent discovery in underserved regions where scouts and elite coaches rarely travel.

In remote areas of Senegal, IOC teams deployed portable tablets to evaluate more than 1,000 youth athletes through standardized physical routines, such as directional jumps, sprints, and push-ups. By comparing captured motion metrics against baseline averages, the AI model flagged positive performance deviations, discovering that nearly 5% of tested participants demonstrated high-performance athletic potential. These young individuals possessed world-class physical capabilities but lacked traditional pathways to discovery simply because of geographic isolation. This mobile scouting approach is equally valuable for geographically expansive nations like Australia, where dispatching scouts across vast, sparsely populated regions is economically unfeasible.

---

[08:31](https://www.youtube.com/watch?v=BFiECev5w7s&t=511s) Focus Areas 3 & 4: Sub-Second Digital Twins and Real-Time Analytics

To enhance both fan engagement and officiating precision, the IOC introduced real-time sports analytics and digital twin systems capable of processing live biomechanical data in under 450 milliseconds—faster than the blink of an eye (200 to 400 milliseconds). These tools convert complex, raw sensor feeds into visual overlays for broadcasters, giving fans instant answers to scoring intricacies, technical difficulty, or athlete identities.

```mermaid flowchart LR accTitle: Real-Time Sports Analytics and Digital Twin Pipeline subgraph Ingestion["Field of Play"] A["Athlete Action / Sensor Capture"] end subgraph Processing["AI Analytics Engine"] B["Biomechanical Analysis & Pose Tracking"] C["Digital Twin Generation (<450 ms)"] end subgraph Distribution["End Users"] D["Broadcasters & Storytelling"] E["Judging & Referees"] F["Fans & Interactive Apps"] end A --> B --> C C --> D C --> E C --> F ```

The system operates seamlessly across multiple disciplines—such as swimming, athletics, and figure skating. In figure skating, it precisely tracks the angle of skate blades during takeoffs and landings, instantly detecting under- or over-rotations that would otherwise require human judges to review up to ten video replays.

During a demonstration at Milano Cortina, a skeptical VIP questioned whether the on-screen digital twin was genuinely real-time or merely replaying pre-recorded animations. The system's fidelity was proven seconds later when American figure skater Ilia Malinin attempted an exceptionally challenging jump and fell; the digital twin captured and rendered the exact failure in sub-second latency, leaving no doubt about the technology's authenticity.

---

[11:20](https://www.youtube.com/watch?v=BFiECev5w7s&t=680s) Trustworthy AI Framework and Governance in Practice

As AI moves into critical operations such as athlete coaching, judging, and fan analytics, establishing institutional trust is paramount. In collaboration with Deloitte, the IOC developed a comprehensive Trustworthy AI Framework based on three practical pillars:

1. **Training & Literacy:** Educating stakeholders across the Olympic ecosystem on safety protocols, data governance, and role-based security. 2. **Seven Pillars of Trust Guidelines:** Clear architectural benchmarks ensuring that all deployed AI applications uphold ethical standards and avoid crossing regulatory or integrity boundaries. 3. **The Practical Governance Checklist:** A functional, accessible checklist (even printable as an Excel sheet) tailored for National Olympic Committees (NOCs) and International Federations (IFs) evaluating vendor pitches.

This checklist protects smaller sports organizations from two major risks when approached by commercial tech providers: rejecting viable innovation out of technical uncertainty (resulting in stagnation), or signing unfavorable agreements that inadvertently violate privacy laws or data storage standards.

---

[13:35](https://www.youtube.com/watch?v=BFiECev5w7s&t=815s) Challenges Encountered and Lessons Learned

Deploying AI across a complex, global organization exposed four key hurdles, each paired with a core lesson learned:

- **Data Readiness vs. Data Foundations First:** While the IOC possesses vast historical archives, much of the data was unstructured, uncleaned, or obsolete. Teams must "sharpen the axe before chopping any trees," as upfront data preparation saves countless hours during model deployment. - **Regulatory Complexity vs. Governance by Design:** Operating across multiple legal jurisdictions with rapidly evolving AI regulations requires embedding compliance from the outset. Organizations must embrace governance proactively rather than responding reactively after a crisis. - **Security Risks vs. Role-Based Access at Scale:** Universal LLM agents create risks of data leakage (such as a newly hired trainee querying executive compensation). Organizations must integrate strict role-based access control (RBAC) into all AI agents. - **Fragmented Ecosystem vs. Inclusive Collaboration:** Managing stakeholders with widely differing levels of AI readiness can slow initial deployment, but engaging both skeptics and technical experts produces robust, universal solutions.

---

[16:17](https://www.youtube.com/watch?v=BFiECev5w7s&t=977s) The Blueprint for AI Transformation

Reflecting on the IOC's broader journey, implementing enterprise AI is compared to driving a car:

```mermaid flowchart LR accTitle: IOC AI Transformation Blueprint subgraph Step1["1. Strategic Focus"] A["Know Destination<br/>(5 Focus Areas & Values)"] end subgraph Step2["2. Tangible Value"] B["Drive & Collect Miles<br/>(Avoid Pointless PoCs)"] end subgraph Step3["3. Guardrails & Trust"] C["Stay on the Road<br/>(Trustworthy AI Framework)"] end Step1 --> Step2 --> Step3 ```

- **Set Strategic Focus:** Define exactly where the organization needs to go using core values and focus areas as an operational filter. - **Deliver Tangible Value:** Avoid accumulating disconnected "proof of concepts" that lead nowhere; focus on shipping production systems that deliver measurable improvements. - **Establish Trust and Guardrails:** Maintain strict governance and continuous reviews to ensure the organization innovates rapidly without compromising ethics or safety.

## Notable quotes > [00:04](https://www.youtube.com/watch?v=BFiECev5w7s&t=4s) whether we like it or not and whether we are ready or not — Alejandro Merino-Madrid

> [00:04](https://www.youtube.com/watch?v=BFiECev5w7s&t=4s) Supporting Athletes, Competition, and Sport

> [00:04](https://www.youtube.com/watch?v=BFiECev5w7s&t=4s) Ensuring Equitable Access

> [00:04](https://www.youtube.com/watch?v=BFiECev5w7s&t=4s) Optimizing Games Operations

> [00:04](https://www.youtube.com/watch?v=BFiECev5w7s&t=4s) Transforming Fan Experiences

> [00:04](https://www.youtube.com/watch?v=BFiECev5w7s&t=4s) Enhancing Internal Efficiency

> [03:04](https://www.youtube.com/watch?v=BFiECev5w7s&t=184s) actually focus on who you are — Maxine Egner

> [13:35](https://www.youtube.com/watch?v=BFiECev5w7s&t=815s) sharpen the axe before chopping any trees — Alejandro Merino-Madrid

## Mentioned Alejandro Merino-Madrid, International Olympic Committee (IOC), Simon Sinek, Olympic AI Agenda, Paris 2024 Olympic Games, Milano Cortina 2026 Winter Games, Maxine Egner, Senegal, Australia, Ilia Malinin, Deloitte, National Olympic Committees (NOCs), International Federations (IFs)

## Video description Alejandro Merino-Madrid - Head of AI, International Olympic Committee (IOC)

What happens when AI not only analyses sport—but actively shapes it in real time? / In this talk, I will show how AI is becoming a system-level force across the Olympic Movement—from protecting athletes against online abuse to optimising Games operations and sustainability at scale. I will explain how we turn ambition into execution through the Olympic AI Agenda and its five strategic focus areas. / You will learn how to move from isolated AI pilots to integrated, high-impact solutions, with real use cases such as digital twins, AI-driven fan experiences, and operational optimisation. I will also show how to balance bold innovation with strong governance—ensuring AI delivers value while safeguarding trust, fairness, and Olympic Values. / Join this session to learn how to turn AI into a true transformation engine for sport—not just another layer of technology.

[Source](https://www.youtube.com/watch?v=BFiECev5w7s)

中文简报

国际奥委会部署AI拦截运动员网暴，将裁判用时压缩至450毫秒内 国际奥委会正在推进AI治理标准化，以便在技术鸿沟跨全球体育界拉大之前，在赛事执裁、移动选材发掘和网络暴力防护方面推广生产级系统。 Alejandro Merino-Madrid 概述了国际奥委会的人工智能战略蓝图，该蓝图围绕五个核心重点领域构建，旨在使技术创新与核心奥林匹克价值观保持一致。他展示了多项切实落地的人工智能应用，包括在2024年巴黎奥运会上保护1.5万多名运动员免受网络骚扰的“AI Sentinel”系统、在塞内加尔部署的基于平板电脑的计算机视觉选材工具，以及用于驱动实时执裁和赛事转播分析、响应时间低于450毫秒的数字孪生系统。为了负责任地治理这一生态，国际奥委会与德勤合作，为各体育单项联合会制定了切实可行的可信AI框架及评估清单。最后，Merino-Madrid 分享了在数据就绪度、前瞻性治理和基于角色的访问控制方面来之不易的经验，敦促各体育机构跳出无休止的概念验证，构建配备严格伦理防线保障的生产级系统。 国际奥委会围绕西蒙·斯涅克（Simon Sinek）的“黄金圈”（Why, How, What）框架构建其AI战略，旨在守护核心奥林匹克价值观的同时，主动引导不可避免的技术变革。 (0:04) 《奥林匹克AI议程》确立了五大战略重点领域：支持运动员、确保公平准入、优化运营、提升观赛体验以及提高内部效率。 (0:04) AI Sentinel 系统通过监测数十亿条社交帖子，并在运动员及其家人看到之前删除针对性的仇恨言论，主动保障运动员的心理健康。 (3:04) 全球体育赛事期间针对个人的网络恶意言论正在迅速上升；尽管相比夏季奥运会总体互动量更低，但2026年米兰-科尔蒂纳冬奥会的相关网络谩骂仍激增了40%以上。 (3:04) 便携式计算机视觉工具通过对塞内加尔等偏远地区的青少年进行能力评估，推动了人才发掘的普惠化，让脱离传统球探网络的高潜力运动员得以被发现。 (6:31)

Reading summary

The International Olympic Committee (IOC) has structured an enterprise AI agenda around five strategic focus areas to actively direct technological disruption toward core Olympic values. By pairing production deployments—such as athlete abuse monitoring, mobile talent scouting, and real-time biomechanics—with institutional governance frameworks, the IOC seeks to modernize Games operations while preventing technological division across sports bodies. Guiding Olympic Transformation Through Strategic Value Filters (0:04) Drawing on Simon Sinek's Golden Circle framework, the IOC structured its strategy around the premise that artificial intelligence will inevitably transform the sports landscape. Rather than waiting passively for market forces to shape the industry, the organization established five strategic filters to prioritize internal development and capital allocation: supporting athletes and competition, ensuring equitable technological access, optimizing Games operations, transforming fan experiences, and enhancing internal efficiency. These focus areas function as an operational North Star to align innovation with foundational Olympic values of fairness, excellence, and solidarity. By applying these criteria, the IOC aims to bypass disconnected proof-of-concept experiments and direct funding strictly toward deployed production systems that deliver measurable improvements. Deploying AI Sentinel Against Targeted Online Abuse (3:04) To counter targeted toxicity, hate speech, and harassment directed at athletes, the IOC developed an AI Sentinel system to scan social media platforms and flag abusive content before competitors or their families encounter it. During the Paris 2024 Olympic Games, the system monitored billions of posts, identifying 2.4 million messages referencing athletes, officials, or teams, and flagged over 10,000 hate speech messages for immediate removal. The initiative protected 15,000 athletes across both Summer and Winter Games and provided relief to 353 competitors actively targeted by coordinated cyber-harassment campaigns. The IOC observed that online hostility does not track overall audience size. At the Milano Cortina 2026 Winter Games, despite a 20% decline in general engagement compared to summer editions, online abuse surged by more than 40%. Competitors including Botswana's Paris 2024 flagbearer, Maxine Egner, reported that automated content filtering shielded their mental well-being, allowing them to step away from toxicity and concentrate on athletic performance. Democratizing Athletic Scouting in Underserved Regions (6:31) To prevent advanced analytics from widening the competitive gap between wealthy sports organizations and resource-constrained bodies, the IOC introduced computer-vision screening tools to democratize talent discovery in regions lacking traditional scouting networks. In remote areas of Senegal, field personnel used portable tablets to assess more than 1,000 youth athletes performing standardized drills such as directional jumps, sprints, and push-ups. By comparing captured motion data against baseline athletic benchmarks, the AI model identified that nearly 5% of participants demonstrated high-performance athletic potential despite having no previous access to elite coaching or scouting. The IOC noted this mobile approach offers a cost-effective scouting framework for expansive territories such as Australia, where dispatching talent scouts across vast, sparsely populated regions is economically unfeasible. Real-Time Biomechanics and Sub-Second Digital Twins (8:31) To enhance broadcast storytelling and refereeing precision, the IOC implemented real-time digital twin architectures capable of processing live biomechanical data in under 450 milliseconds—faster than a human eye blink of 200 to 400 milliseconds. Operating across sports like swimming, athletics, and figure skating, the engine converts raw sensor feeds and pose tracking into live visual overlays. In figure skating, the system tracks blade angles on takeoffs and landings to detect rotation errors that previously required judges to review up to ten video replays. The sub-second latency of the technology was validated during a demonstration at Milano Cortina when American figure skater Ilia Malinin fell during a complex jump. The digital twin captured and rendered the missed landing in under 450 milliseconds, demonstrating that the visual output was an authentic real-time calculation rather than a pre-rendered graphic. Institutional Trust Frameworks and Enterprise Implementation Bottlenecks (11:20) To regulate critical AI applications, the IOC partnered with Deloitte to design a Trustworthy AI Framework based on literacy training, Seven Pillars of Trust architectural benchmarks, and a practical governance checklist. This checklist is designed to help National Olympic Committees and International Federations evaluate commercial technology pitches without either stalling out of technical hesitation or signing agreements that violate data privacy and legal standards. Deploying these systems surfaced four operational lessons across the IOC's global ecosystem: vast historical archives required extensive upfront data cleaning ('sharpening the axe'); cross-border operations demanded proactive legal compliance embedded from the design phase; internal large language model implementations required role-based access controls to prevent data leakage; and varying technical maturity across stakeholder groups required early collaboration with both technical experts and skeptics. The IOC frames its AI transformation as an ongoing operational balance between technological ambition and ethical guardrails, prioritizing production value over exploratory pilots. However, maintaining consistent data governance and equitable access across an international ecosystem of independent federations with sharply divergent technical and financial capacities remains an unresolved implementation challenge.

精炼阅读摘要

国际奥林匹克委员会（IOC）认为，人工智能对全球体育产业的重塑是不可逆转的必然趋势，体育组织必须主动引导技术演进以契合公平、卓越与团结等核心价值观。为此，IOC确立了涵盖运动员保护、技术普惠、赛事运营、观众体验与内部效能的AI转型蓝图，致力于将前沿算法转化为具备实际生产力与伦理约束的系统性能力。 AI Sentinel拦截网络仇恨保护运动员 (3:04) 针对参赛选手日益面临的定向网络攻击、仇恨言论及基于种族、宗教、国籍或政治立场的骚扰，IOC开发并部署了名为“AI Sentinel”的智能监控系统。该系统能够在各大社交平台上扫描并过滤海量公开帖文与评论，在运动员及其亲友接触到有害内容之前，自动识别并提交平台实施封禁处理。在巴黎2024年奥运会期间，AI Sentinel在数十亿条监测数据中筛查出240万条直接提及运动员、官员或团队的信息，从中标记并立即提交删除了超过1万条明确的仇恨言论帖文；该服务在夏奥会与冬奥会期间累计为15,000名运动员提供了数字化防护，并重点支援了353名遭受有组织网络攻击的选手，提供了关键的心理减压支持。 在米兰-科尔蒂纳2026年冬奥会的相关测试中，尽管冬季赛事的社媒互动总量通常比夏季赛事低约20%，但网络辱骂与仇恨言论的发生率却逆势激增超过40%，进一步印证了自动化防护工具的必要性。巴黎2024年奥运会博茨瓦纳代表团旗手Maxine Egner指出，选手在遭遇大量仇恨信息和针对性图像攻击时承受巨大精神负担，而IOC的网络防护系统能帮助运动员隔绝恶劣舆论环境，维护心理健康，从而将注意力重新聚焦在赛场竞技与自我发挥之上。 计算机视觉与便携评估下沉选材 (6:31) 如果缺乏主动干预，先进AI技术的集中往往会拉大资金雄厚机构与资源匮乏组织之间的竞争差距。为防止出现技术鸿沟，IOC正通过计算机视觉与运动机能评估模型推动体育选材的平权化。在传统球探和顶级教练极少涉足的偏远地区，IOC技术团队借助便携式平板电脑采集基层青少年运动员的标准化体能测试数据，打破以往依赖昂贵专业实验室的选拔门槛。 在塞内加尔的偏远地区，IOC团队利用搭载AI评估模型的移动终端，对1,000多名当地青年进行了定向跳跃、冲刺和俯卧撑等基础体能测试。通过将实时动作捕捉数据与基准平均线进行比对，模型发现近5%的受测青少年展现出符合高水平竞技要求的身体潜能偏差。根据IOC的评估，这些群体虽然具备世界级的身体机能素质，但过去受限于地理区位阻隔而完全缺乏被职业体系发掘的渠道。此外，这种移动端轻量化筛选方案同样适用于澳大利亚等地域辽阔的国家，有效解决了在人口稀疏地区派遣实地球探经济成本过高的问题。 亚秒级数字孪生提升转播与判罚精度 (8:31) 为提高赛事转播的信息表现力与裁判裁决的客观度，IOC引入了具备亚秒级响应能力的实时体育分析与数字孪生系统。该系统能够在450毫秒以内完成现场生物力学传感器数据的采集、姿态追踪与三维模型重建，其运算速度快于人类正常眨眼所需的200至400毫秒。系统生成的动态可视化图层可即时供给赛事转播商，协助解说员向受众实时解析复杂的计分规则、动作技术难度与选手即时状态。 该技术已拓展至游泳、田径和花样滑冰等多项运动中。以花样滑冰为例，算法能精确追踪运动员起跳与落冰时冰刀的切入角度，瞬间识别周数不足或过度旋转，替代以往裁判往往需要反复调取多达十次慢动作回放的裁量流程。在米兰-科尔蒂纳冬奥会的一次现场技术演示中，面对数字孪生是否仅为预录动画的现场质疑，美国花滑选手Ilia Malinin在尝试超高难度跳跃时发生失误摔倒，数字孪生系统以亚秒级延迟完整复刻并实时渲染了其失误轨迹，直接验证了该技术在严苛竞赛环境下的真实保真度。 可信治理框架与规模化落地的实施挑战 (11:20) 随着AI逐步深入执教辅助、裁决判定及受众数据分析等核心场景，建立机制化的制度信任成为推广前提。IOC与Deloitte联合制定了“可信AI框架”（Trustworthy AI Framework），主要由三大支柱构成：面向奥林匹克生态全员的安全协议、数据治理与分级安全培训；规范算法部署、防止触碰伦理与诚信红线的“七大信任支柱”准则；以及专为各国家奥委会（NOC）和国际单项体育联合会（IF）量身定制的实用治理审查清单。这一清单甚至可以简化为Excel表格形式，协助缺乏技术背景的中小型体育组织在评估商业供应商时，既避免因技术不确定性而盲目抗拒创新，也防止因草率签约而触犯隐私法规或数据合规红线。 在推动AI跨全球多机构落地的过程中，IOC总结出四项关键挑战与对策：首先，面对历史档案庞大但多为非结构化、未清洗脏数据的现实，必须秉持“磨刀不误砍柴工”的原则，优先夯实底层数据质量；其次，面对跨国运营所涉及的差异化法案，推行“设计即合规”（Governance by Design）的前置防护；再次，为防范通用大模型代理带来的数据泄露风险（例如新入职员工越权查询高管薪酬），全面集成严格的基于角色的访问控制（RBAC）；最后，在技术准备度参差不齐的生态中坚持包容性协作，融合技术专家与审慎者的意见，以此避免陷入孤立且无法投入生产的无效概念验证（PoC）。 IOC将体育领域的企业级AI推进比作驾驶汽车：既需要以核心价值观与五大重点领域锚定方向，也必须聚焦能够持续输出可衡量价值的生产级系统，同时依靠可信治理框架充当安全护栏。然而从其总结的经验可以看出，跨国界的监管差异、成员机构间悬殊的数据准备度以及技术鸿沟，仍是这套转型蓝图在全生态落地过程中必须持续应对的结构性制约。

Detailed notes — zh

## Background 该内容由 Ai4 于 2026 年 8 月 21 日发布在 YouTube 平台。页面简介显示，国际奥委会（IOC）人工智能负责人 Alejandro Merino-Madrid 在本次演讲中阐述了 AI 如何成为席卷奥林匹克运动的系统级力量，深入剖析了“奥林匹克 AI 议程”的五大战略支柱，并分享了从保护运动员、优化赛事运营到数字孪生与可信 AI 治理的落地实践。 ## TL;DR 国际奥委会通过“奥林匹克 AI 议程”确立了覆盖支持运动员、技术普惠、赛事运营、观赛体验与内部效能的五大战略支柱。在落地实践中，IOC 部署了用于实时拦截全网仇恨言论的“AI 哨兵”系统，借助便携平板电脑在塞内加尔偏远地区公平选材，并实现了延迟低于 450 毫秒的赛场实时数字孪生分析。为平衡技术创新与伦理风险，IOC 联合德勤推出了包含七大支柱与落地清单的“可信 AI 框架”，总结了数据地基先行、合规前置等经验。最终，演讲以“驾车前行”为喻，提炼出兼顾战略聚焦、切实价值与安全护栏的组织 AI 转型蓝图。 ## Atmosphere & style 整场演讲呈现出务实严谨、前瞻宏大又兼具人文关怀的专业基调。演讲者以黄金圈逻辑层层推导，在讲述运动员遭受网暴与塞内加尔青年选材时饱含共情与使命感，而在拆解毫秒级运动捕捉和纸质核对清单时又展现出极强的工程落地思维。真实赛场案例与生动形象的比喻交织，让宏大的国际体育治理框架变得极为接地气且富有感染力。 ## Key takeaways - [00:04](https://www.youtube.com/watch?v=BFiECev5w7s&t=4s) 国际奥委会（IOC）从黄金圈法则出发主动制定奥林匹克 AI 议程，旨在将不可阻挡的技术浪潮转化为践行公平、卓越与团结等核心价值观的内生动力。 - [01:44](https://www.youtube.com/watch?v=BFiECev5w7s&t=104s) IOC 确立了支持运动员、技术普惠、优化赛事组织、重塑观众互动及提升内部效率五大战略重点，作为所有 AI 项目立项与预算支持的硬性战略过滤器。 - [03:04](https://www.youtube.com/watch?v=BFiECev5w7s&t=184s) 通过部署“AI 哨兵”系统实时监测全网社交媒体互动，在当事人察觉前拦截并删除针对运动员的网络暴力与仇恨言论，在巴黎奥运会成功清除逾万条恶意信息。 - [03:04](https://www.youtube.com/watch?v=BFiECev5w7s&t=184s) 数据显示冬奥会网络霸凌与攻击言论较夏奥会激增超 40%，证明在赛事不同阶段持续提供网络暴力防御与运动员心理支持具有极高的紧迫性。 - [06:31](https://www.youtube.com/watch?v=BFiECev5w7s&t=391s) 利用便携式平板与 AI 动作分析深入塞内加尔偏远乡村，成功发掘近 5% 的潜在高水平体育幼苗，大幅降低球探选拔门槛并促进全球体育资源普惠。 - [08:31](https://www.youtube.com/watch?v=BFiECev5w7s&t=511s) 打造低于 450 毫秒超低延迟的实时数字孪生系统，可在眨眼之间精准还原花样滑冰等项目中的微观角度偏差与动作失误，革新专业裁判与观众观赛体验。 - [11:20](https://www.youtube.com/watch?v=BFiECev5w7s&t=680s) 携手德勤推出“可信 AI 框架”与实操核对清单，帮助中小型体育组织在面对科技伙伴方案时明确数据安全与合规边界，避免盲目跟风或错失发展良机。 - [13:35](https://www.youtube.com/watch?v=BFiECev5w7s&t=815s) 数据地基先行（Data Foundations First）：前期在非结构化与过时数据清洗治理上投入的每一小时，都能为后续规模化应用开发节省数十倍时间成本。 - [13:35](https://www.youtube.com/watch?v=BFiECev5w7s&t=815s) 应对全球多变监管需将治理前置（Governance by Design），并在大模型与智能体中严格内建基于角色权限的安全控制（RBAC），防止敏感信息越权泄露。 - [16:17](https://www.youtube.com/watch?v=BFiECev5w7s&t=977s) AI 转型遵循三步法则：一是明确战略方向与价值观导航，二是踩下油门摆脱无效概念验证（PoC）交付实际价值，三是系紧安全带用信任框架筑牢安全护栏。 ## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=BFiECev5w7s&t=4s) 为什么奥林匹克运动需要 AI 议程：从黄金圈法则出发 国际奥委会（IOC）人工智能负责人 Alejandro Merino-Madrid 引用了 Simon Sinek 著名的“黄金圈（Golden Circle）”思考框架——从“为什么（Why）”出发，延伸至“如何做（How）”与“做什么（What）”，最后在“那又怎样（So What）”中总结经验教训与转型蓝图。 谈及为什么奥林匹克运动必须设立 AI 议程，Merino-Madrid 指出：“无论我们喜欢与否，无论我们是否做好准备，AI 浪潮都必将到来。”与其被动观望并寄希望于变革自然向好，IOC 选择主动出击，引导这场席卷整个体育产业的技术变革。IOC 的雄心是通过 AI 积极赋能并践行公平（fairness）、卓越（excellence）和团结（solidarity）等核心奥林匹克价值观，全面提升奥运体验。 ### [01:44](https://www.youtube.com/watch?v=BFiECev5w7s&t=104s) 推进转型的五大核心战略领域 为了将愿景转化为具体的执行路径，IOC 确立了五大战略重点领域。这五大领域不仅是行动的“北极星”，更是一个严格的战略过滤器——任何脱离这五个维度的项目，IOC 都不会予以预算支持或启动立项。 ```mermaid flowchart TD accTitle: IOC 五大 AI 战略重点领域 subgraph Core["以人为本与生态公平"] A["1. 支持运动员、比赛与竞技"] B["2. 促进 AI 普惠共享，凝聚行业"] end subgraph Operation["赛事与受众体验"] C["3. 高效、可持续的赛事组织"] D["4. 升级全球奥运迷互动体验"] end subgraph Internal["组织内生动力"] E["5. 提升 IOC 内部运营效率"] end Core --> Operation --> Internal ``` - **领域一（支持运动员与竞技）**：AI 的首要任务是服务于运动员、比赛公平性与体育运动本身。 - **领域二（促进技术普惠）**：新技术既能制造数字鸿沟，也能促进团结。IOC 致力于消除资源壁垒，让 AI 红利惠及全球体育产业各方。 - **领域三（优化赛事组织）**：借助 AI 让奥运会的筹备与运营变得更快、更高效、更具可持续性。 - **领域四（重塑观众体验）**：利用 AI 打造前所未有的观赛互动方式，触达并吸引更广泛的新一代年轻受众。 - **领域五（提升内部效率）**：在赋能外部生态的同时反求诸己，提升 IOC 自身的日常工作与决策效率。 ### [03:04](https://www.youtube.com/watch?v=BFiECev5w7s&t=184s) 守护运动员：利用 AI 哨兵抵御网络暴力与仇恨言论 竞技场上的运动员背负着常人难以想象的心理压力。Merino-Madrid 描绘了一个揪心的场景：一名运动员苦练一生终于站上奥运舞台，但在出战前夕却发现家人在暗自垂泪。追问之下才得知，互联网上正充斥着针对该运动员的恶毒网络攻击，涉及种族、宗教、肤色、政治观点等各种人身侮辱与威胁。 为了彻底清除这种在体育界绝不容忍的霸凌行径，IOC 部署了专门的“AI 哨兵（AI Sentinel）”系统，对全网数十亿条社交媒体帖子和评论进行实时监测。一旦发现针对运动员或团队的仇恨言论，系统会在当事人及其家属看到之前直接向社交平台举报并予以删除，同时为受波及的运动员提供及时的心理支持。 - **巴黎 2024 奥运会数据**：系统筛查了数十亿条互动，识别出 240 万条直接提及运动员、官员或队伍的评论，并精准定位并清除了超过 10,000 条恶性仇恨言论。 - **保护规模**：在最近的两届夏季与冬季奥运会中，该系统累计保护了 15,000 名运动员，其中 353 名遭受集中网络攻击的选手得到了重点防护。 - **米兰-科尔蒂纳 2026 趋势**：冬奥会虽然整体互动量比夏奥会低约 20%，但网络霸凌和网络暴力言论反而激增了 40% 以上。 - **运动员亲历**：巴黎奥运会博茨瓦纳旗手 Maxine Egner 现身说法，表示自己在参赛期间遭遇了铺天盖地的恶意图片与帖文轰炸，正是 IOC 的网络暴力防护服务帮助她过滤掉了恶意攻击，“让你能够专注于自己的运动，专注于做真正的自己”。 ### [06:31](https://www.youtube.com/watch?v=BFiECev5w7s&t=391s) 普惠共享：用平板与 AI 在偏远地区发掘未来之星 AI 如果任由市场自由发展，往往会拉大资源雄厚组织与落后群体之间的差距。为践行“团结”理念，IOC 积极推动 AI 工具普惠化。两年前，IOC 团队携带普通平板电脑深入塞内加尔的偏远山村与深谷，去探寻那些因为地处偏远、缺乏球探踏足而被埋没的体育幼苗。 测试过程中，当地 1,000 多名青少年在镜头前完成标准动作（如左右横跳、5 次俯卧撑、垂直纵跳、全力冲刺等）。AI 模型实时比对运动生物力学指标，自动捕捉正向偏离平均值的“绿色信号（Green Flags）”。测试结果令人震惊：在这批从未接受过专业训练的孩子中，有将近 5% 的人展现出了成为高水平精英运动员的惊人天赋。 这项技术不仅改变了塞内加尔偏远青年的命运，同样吸引了像澳大利亚这样资源丰富的大国。澳大利亚同样面临人口分布稀疏、无法在每个角落常驻球探的痛点，便携式 AI 选材方案为全球各地区提供了极具成本效益的公平选拔机会。 ### [08:31](https://www.youtube.com/watch?v=BFiECev5w7s&t=511s) 实时体育分析与数字孪生：在眨眼之间呈现赛场洞察 在提升观众体验方面，IOC 为持权转播商和普通车迷提供了实时体育数据分析工具。系统将过去晦涩枯燥的传感器数据瞬间转化为直观的可视化图表与专业解说素材。观众无论是坐在电视机前还是身处场馆，都可以通过移动端在半秒内获取专家级的实时裁决解析。 系统在花样滑冰项目中甚至能实时捕捉冰刀与冰面的微小夹角，精准判断选手在起跳与落冰时是否存在周数不足或过度旋转。通常需要裁判肉眼回放十次回放才能定性的动作，AI 在瞬间即可完成判定。更关键的是处理速度：人类一次眨眼的时间大约是 200 到 400 毫秒，而该技术从赛场动作发生到观众指尖呈现，延迟低于 450 毫秒。 在米兰-科尔蒂纳冬奥会展示该系统时，曾有贵宾怀疑屏幕上的“数字孪生（Digital Twin）”只是预录好的动作动画库匹配。恰巧在随后的比赛中，顶级花样滑冰名将 Ilia Malinin 在挑战超高难度跳跃动作时出现罕见失误摔倒，系统毫秒不差地在数字孪生上同步还原了这一失误轨迹，以无可辩驳的真实性彻底打消了所有人的疑虑。 ### [11:20](https://www.youtube.com/watch?v=BFiECev5w7s&t=680s) 可信 AI 框架：七大支柱与落地实操清单 随着 AI 全面渗透到运动员训练、裁判判决和观众互动等核心业务，建立技术信任成为当务之急。为此，IOC 联合德勤（Deloitte）共同研发了“可信 AI 框架（Trustworthy AI Framework）”，该体系由三大核心组件构成： 1. **普及培训（Training）**：提升全员数字素养，帮助不同岗位的从业者理解角色安全、系统防御等基础概念。 2. **七大可信支柱指南（Guidelines - 7 Pillars of Trust）**：明确 AI 落地不可逾越的红线与伦理规范。 3. **实操核对清单（Practical Checklist）**：一张清晰明了的纸质/Excel 检查清单。 Merino-Madrid 特别强调，在充斥着复杂智能体和华丽仪表盘的今天，最管用的往往是一张可以打印出来、用笔逐项打勾的 Excel 清单。当中小型国家奥委会（NOC）或国际单项体育联合会（IF）面对科技巨头开出的诱人合作方案时，往往面临两难：因看不懂而盲目拒绝，会痛失发展良机；因盲目跟进而匆忙签约，又可能在不知情中侵犯数据隐私、跨越合规红线。通过这张清单，管理人员能直接与技术伙伴对话，逐项核查数据存储位置与合规机制，确保在安全边界内大胆创新。 ### [13:35](https://www.youtube.com/watch?v=BFiECev5w7s&t=815s) 挑战与教训：从数据地基到生态协同 回顾 IOC 的 AI 探索之路，Merino-Madrid 将遇到的四大核心挑战与对应的经验教训进行了深度复盘： | 挑战（Challenge） | 经验教训（Lessons Learned） | 核心要点与对策 | |:--- |:--- |:--- | | **数据尚未就绪**<br>数据体量庞大但非结构化、不清晰、过时 | **数据地基先行**<br>(Data Foundations First) | 磨刀不误砍柴工。在清洗、结构化和治理数据上投入的每一小时，都会在后续开发中节省数十倍时间。 | | **全球监管复杂性**<br>各国家与地区法律动态多变且差异巨大 | **架构内建合规治理**<br>(Governance by Design) | 治理必须前置。企业要么在项目一开始就构筑合规体系，要么在遭遇危机后被迫补救，必须做前者。 | | **大模型安全与越权风险**<br>同个智能体可能泄露高管薪酬等敏感信息 | **基于角色的大规模安全控制**<br>(Security at Scale) | 严格嵌入基于角色权限访问控制（RBAC），确保不同层级员工在使用同一 AI 时获取匹配的信息。 | | **高度碎片化的生态系统**<br>各利益相关方的技术素养与接受度参差不齐 | **生态协同是破局关键**<br>(Collaboration is Key) | 将怀疑论者与狂热支持者、技术小白与资深专家聚在同一张桌子上，共同打磨出让所有人满意的终极方案。 | ### [16:17](https://www.youtube.com/watch?v=BFiECev5w7s&t=977s) AI 转型蓝图：开车前行的三步法则 在演讲尾声，Merino-Madrid 用“驾驶汽车”生动形象地总结了组织实现 AI 转型的三步蓝图： - **第一步：明确前行方向（战略聚焦 / Strategic Focus）** 上车前必须明确目的地。对 IOC 而言，五大战略重点和奥林匹克价值观就是唯一的导航路线，绝不盲目偏航。 - **第二步：踩下油门交付切实价值（Deliver Tangible Value）** “世界上最好的战略如果不去落地执行，甚至比不上承载它的那张白纸的价值。”组织必须真正上路积累里程，拒绝停留在无法规模化推广的概念验证（PoC）玩具项目中。 - **第三步：系好安全带，切莫翻车（建立信任与护栏 / Trust & Guardrails）** 一旦车辆高速行驶，必须依靠完善的可信 AI 治理框架设立清晰护栏，并在行进过程中持续审查纠偏，确保创新始终行驶在安全合规的轨道之上。 ## Notable quotes > [00:04](https://www.youtube.com/watch?v=BFiECev5w7s&t=4s) 无论我们喜欢与否，无论我们是否做好准备，AI 浪潮都必将到来。 — Alejandro Merino-Madrid > [01:44](https://www.youtube.com/watch?v=BFiECev5w7s&t=104s) 任何脱离这五个维度的项目，IOC 都不会予以预算支持或启动立项。 — Alejandro Merino-Madrid > [03:04](https://www.youtube.com/watch?v=BFiECev5w7s&t=184s) 让你能够专注于自己的运动，专注于做真正的自己 — Maxine Egner > [06:31](https://www.youtube.com/watch?v=BFiECev5w7s&t=391s) 在这批从未接受过专业训练的孩子中，有将近 5% 的人展现出了成为高水平精英运动员的惊人天赋。 — Alejandro Merino-Madrid > [08:31](https://www.youtube.com/watch?v=BFiECev5w7s&t=511s) 人类一次眨眼的时间大约是 200 到 400 毫秒，而该技术从赛场动作发生到观众指尖呈现，延迟低于 450 毫秒。 — Alejandro Merino-Madrid > [11:20](https://www.youtube.com/watch?v=BFiECev5w7s&t=680s) 最管用的往往是一张可以打印出来、用笔逐项打勾的 Excel 清单。 — Alejandro Merino-Madrid > [13:35](https://www.youtube.com/watch?v=BFiECev5w7s&t=815s) 在清洗、结构化和治理数据上投入的每一小时，都会在后续开发中节省数十倍时间。 — Alejandro Merino-Madrid > [13:35](https://www.youtube.com/watch?v=BFiECev5w7s&t=815s) 企业要么在项目一开始就构筑合规体系，要么在遭遇危机后被迫补救，必须做前者。 — Alejandro Merino-Madrid > [13:35](https://www.youtube.com/watch?v=BFiECev5w7s&t=815s) 将怀疑论者与狂热支持者、技术小白与资深专家聚在同一张桌子上，共同打磨出让所有人满意的终极方案。 — Alejandro Merino-Madrid > [16:17](https://www.youtube.com/watch?v=BFiECev5w7s&t=977s) 世界上最好的战略如果不去落地执行，甚至比不上承载它的那张白纸的价值。 — Alejandro Merino-Madrid ## Mentioned Alejandro Merino-Madrid, 国际奥委会（IOC）, Simon Sinek, 黄金圈法则（Golden Circle）, 奥林匹克 AI 议程（Olympic AI Agenda）, AI 哨兵（AI Sentinel）, 巴黎 2024 奥运会, Maxine Egner, 米兰-科尔蒂纳 2026 冬奥会, 塞内加尔, 澳大利亚, Ilia Malinin, 德勤（Deloitte）, 可信 AI 框架（Trustworthy AI Framework）, 国家奥委会（NOC）, 国际单项体育联合会（IF） ## Video description Alejandro Merino-Madrid - Head of AI, International Olympic Committee (IOC) What happens when AI not only analyses sport—but actively shapes it in real time? / In this talk, I will show how AI is becoming a system-level force across the Olympic Movement—from protecting athletes against online abuse to optimising Games operations and sustainability at scale. I will explain how we turn ambition into execution through the Olympic AI Agenda and its five strategic focus areas. / You will learn how to move from isolated AI pilots to integrated, high-impact solutions, with real use cases such as digital twins, AI-driven fan experiences, and operational optimisation. I will also show how to balance bold innovation with strong governance—ensuring AI delivers value while safeguarding trust, fairness, and Olympic Values. / Join this session to learn how to turn AI into a true transformation engine for sport—not just another layer of technology. [Source](https://www.youtube.com/watch?v=BFiECev5w7s)

Source quotes

[0:04] Alejandro Merino-Madrid: 无论我们喜欢与否，无论我们是否做好准备，AI 浪潮都必将到来。 [1:44] Alejandro Merino-Madrid: 任何脱离这五个维度的项目，IOC 都不会予以预算支持或启动立项。 [3:04] Maxine Egner: 让你能够专注于自己的运动，专注于做真正的自己 [6:31] Alejandro Merino-Madrid: 在这批从未接受过专业训练的孩子中，有将近 5% 的人展现出了成为高水平精英运动员的惊人天赋。 [8:31] Alejandro Merino-Madrid: 人类一次眨眼的时间大约是 200 到 400 毫秒，而该技术从赛场动作发生到观众指尖呈现，延迟低于 450 毫秒。 [11:20] Alejandro Merino-Madrid: 最管用的往往是一张可以打印出来、用笔逐项打勾的 Excel 清单。 [13:35] Alejandro Merino-Madrid: 在清洗、结构化和治理数据上投入的每一小时，都会在后续开发中节省数十倍时间。 [13:35] Alejandro Merino-Madrid: 企业要么在项目一开始就构筑合规体系，要么在遭遇危机后被迫补救，必须做前者。 [13:35] Alejandro Merino-Madrid: 将怀疑论者与狂热支持者、技术小白与资深专家聚在同一张桌子上，共同打磨出让所有人满意的终极方案。 [16:17] Alejandro Merino-Madrid: 世界上最好的战略如果不去落地执行，甚至比不上承载它的那张白纸的价值。

Source quotes

[0:04] Alejandro Merino-Madrid: whether we like it or not and whether we are ready or not [0:04] Supporting Athletes, Competition, and Sport [0:04] Ensuring Equitable Access [0:04] Optimizing Games Operations [0:04] Transforming Fan Experiences [0:04] Enhancing Internal Efficiency [3:04] Maxine Egner: actually focus on who you are [13:35] Alejandro Merino-Madrid: sharpen the axe before chopping any trees

Original transcript

[0:04] Good morning everyone. It's a pleasure to be  here today. We're going to get into the Olympic [0:09] AI agenda and I'm going to share with you our  recipe, our blueprint for AI transformation. [0:17] Many of you may be familiar with Simon Synenix  framework start with Y and the golden framework [0:24] the golden circle. We're going to organize  this presentation in a similar way. [0:28] We will start with why. Why does the  Olympic movement need a AI agenda? [0:33] Continue with how. How are we advancing towards  this ambition and what exactly are we doing? [0:40] In the so what section we will  look back at the AI journey [0:44] and the challenges, the lessons learned we had  as well as this blueprint for AI transformation. [0:54] Let's start with why. Why do we need  a AI agenda for the Olympic movement? [0:59] I believe AI is going to happen whether we  like it or not and whether we are ready or not. [1:05] And we decide to be ready and to drive  this change. Since the change is going [1:10] to happen for the sports industry, we want  to make sure that we land in the right place [1:14] according to our Olympic values and our ambition.  So we want to actively drive this change [1:21] rather than wait and hope that this change is for  the best. With that in mind, this is our ambition. [1:28] The International Olympic Committee inspires and  accelerates change, leveraging AI to advance the [1:37] Olympic values, fairness, excellence, solidarity,  and the Olympic experience. This is our goal. [1:44] Now, how are we advancing this ambition? How are  we advancing this goal? We have five focus areas. [1:51] Think of them as our northern star where  we want to go. Or perhaps it's a filter. [1:57] Everything outside of here is not our business.  We are not budgeting it. We are not starting it. [2:02] We believe that AI must first support the  athletes. So our first focus area is all about [2:08] supporting the athletes, competition and sport.  Second, AI like any new technology has the power [2:15] to divide or to unite. We want to unite and for  doing that the benefits of AI must be available [2:22] for everyone within the sports industry. The third  focus area is about what we do best organizing the [2:28] games. We want to organize them faster, more  efficiently, more sustainably thanks to AI. [2:36] The fourth focus area is hopefully about  all of you Olympic fans. We want to give you [2:42] a yet better experience, new experiences, engage  you in new ways and even reach newer fans. [2:48] And our fifth focus area after looking  everywhere within the Olympic movement, [2:53] we also want to look within and  improve our internal efficiency. [2:59] Now, what exactly are we doing with AI? [3:04] As I was saying, first AI must support the  athletes. And I want you for a second to imagine. [3:09] Imagine you are an athlete. You have been training  your whole life and today is the day. Today you [3:15] are competing in the Olympic games. Before that  you see your spouse or your partner or your child [3:23] or parent and they are frustrated. Perhaps they're  even crying. So you ask, "Hey, what's going on?" [3:29] They tell you, "Don't worry. It's just  something is on the internet." And [3:32] you worry. So you go to the internet. In the  internet, there is a campaign against you [3:37] and it's full of hate speech. They are  discriminating you, threatening you, insulting you [3:42] based on your race, religion, ethnicity, the  color of your skin, political views or anything [3:47] else that makes you an individual. This is not  fair and this has no place in the world of sport. [3:52] AI can help and this is our answer. We develop  a AI sentinel that parses billions of posts and [3:59] comments. Let me show the numbers for Paris 24.  It checks billions with V of posts and comments [4:06] identifying 2.4 millions that directly name  or address an athlete, an official or a team. [4:14] Out of these, most are good comments. I like how  he plays. I don't like how he jumps. I don't like [4:20] how he runs. Whatever. That's all fine. But  10,000, more than 10,000 are not fine. Are [4:26] hate speech. He hate speech has no place. So the  Sentinel will report them to the social networks [4:33] and remove them before most more likely than not  before the athletes and the families can see them. [4:40] Over the last two Olympic games, summer and  winter, 15,000 athletes were protected with this [4:46] tool. Especially 353 were targeted by different  campaigns and particularly benefited from this [4:53] sealing. And if you're thinking, well, these  are the numbers from Paris. It's two years ago. [4:58] Perhaps we have advanced as a society.  We are better now. If you think that, [5:02] I have bad news for you. The numbers below  are for Milano Cortina just a few months ago. [5:07] Even Winter Olympic Games has 20% less  engagement. The hate speech, the abusive [5:13] speech, the cyber abuse was more than 40% more.  This tool is now more important than ever. [5:20] Let's hear from Maximar. She was the  flag bearer for Botswana in Paris 2024. [5:28] My name is Maxine Edgar. I was a flag bearer  for Botswana at the Paris 2024 Olympic Games [5:35] and I got a bombardment of messages. There  was one specific image that was being sent [5:41] around and there were horrible  posts about me on the internet. [5:46] The Olympic AI agenda in its focus area one is all  about using technology to protect the athletes. [5:55] The cyber abuse protection service that we offer  at the Olympic Games brings that vision to life. [6:01] Across Paris 2024 and Milano Cortina 2026, the  system analyzed millions of social media posts [6:10] and we were able to provide psychosocial support  to those athletes affected by the online abuse. [6:17] With the service being offered by the ILC, [6:20] it helped you focus on your sport  and actually focus on who you are. [6:31] The second focus area has to do with making  sure the benefits of AI are available everywhere [6:37] for everyone. AI is a relatively new technology  in sports and it has the power to divide or the [6:44] power to unite. If let alone it can divide  because those that have the best resources, [6:50] the most AI literacy, the most money, they will  acquire this AI solutions and the gap with those [6:56] that do not have access to it will increase. Our  job is to do exactly the opposite. We want the [7:02] benefits of AI to be available for everyone.  Let me play this video while I explain a bit. [7:08] Two years ago, the AOC went to Seneagal  to particularly remote areas of Seneagal, [7:14] up in a mountain, behind a valley, whatever it  was. There were lots of talent there, but not lots [7:19] of opportunities because coaches, head hunters do  not reach there. Simply, they don't live there. [7:25] With AI, we can unear part of this talent.  What we did was go with a tablet and test [7:32] over 1,000 young athletes performing predefined  routines. Think of it as jump left, jump right, [7:39] do five push-ups, jump as high as you can, sprint.  The AI model was comparing metrics with an average [7:46] and detecting green flags, positive deviations.  In fact, almost 5% of these over thousand athletes [7:54] had the potential to change their life and  the life of their communities through sport. [7:58] They had the potential to be high performance  athletes. They were just born in the wrong [8:03] in the wrong place to be detected. AI can  help us go anywhere. And if you're thinking [8:08] this is interesting for Seneagal but not for  every country, we are speaking with countries as [8:12] developed of as Australia as rich as Australia.  They also need something like this because [8:17] they have also unpopulated areas and it's not  economically wise to send head hunters everywhere. [8:23] So this is a solution for bringing  the benefits of AI really to everyone. [8:31] We were speaking about engaging fans and giving  a new experience. This is a tool we have for that [8:37] real time sports analytics. We give this  tool to our broadcasters. So invisible data, [8:44] boring data transforms into insights and  with the right angle and the right comment, [8:49] it transforms into unforgettable stories.  We can also give it to fans directly. [8:54] you are home watching or maybe in the stadium and  even you're watching the action. You want to have [8:59] an expert at your fingertips. Hey, why was that  scored that way or why is that that difficult [9:04] or who is this guy actually? All of that can be  answered and it takes less than half a second [9:10] from the screen to your fingertips. So, you  have realtime information judging and referring [9:15] you have seen that it was detecting even  the angle of the blades for a figure skater [9:19] when he or she jumps and lands. Is there a under  rotation, a over rotation? The human eye without [9:26] aid will need 10 repetitions to identify certain  penalties. AI can do it seamlessly in no time. [9:34] That was figure skating. But why to stop  there? It is multisport. Athletics, swimming, [9:40] synchronized jumping, whatever it is, we can  deploy AI that helps that particular sport. [9:46] It is multi-purpose. We mentioned  judging a referring or fun engagement, [9:50] but how about giving athletes insights of how  they perform, why they won or why they didn't. [9:57] Trust is embedded in the system  is the most important part for us. [10:01] And all of that is supported by a technology that  acts in less than half a second. The blink of an [10:06] eye is between 200 and 400 milliseconds. With this  technology, what happens in the field of play, [10:11] it takes less than 450 milliseconds to be  at your fingertips. the blink of an eye. [10:20] When we were, allow me to  present a small anecdote. [10:24] When we were in Milano Cortina presenting this  solution, we had a VIP looking at his screen [10:29] and thinking, well, maybe this is true,  maybe, maybe it is not. So, he asked us, [10:35] is this digital twin of an athlete a real thing?  or it's like a predefined pre-recorded set of [10:41] routines and they just I don't know identify  the move and show you something similar. [10:46] We're telling him no, no, believe us. This is  technology you can trust. This is the real thing. [10:52] I'm not sure if he believed me, but for  sure he believed what happened next. [10:56] Maybe you know Ilia Molinin is perhaps the  most spectacular figure skater of all times [11:02] in Milano Cortina. He did not have his best day.  He attempted a particularly challenging jump [11:08] and he failed. And of course,  that's what the digital twin showed. [11:11] Exactly what happens on the field of play at  your fingertips in less than half a second. [11:20] We have deep dive in three  solutions of what can we do with AI? [11:25] But the bigger question is can we trust AI? [11:28] AI is not staying at the edges. It's going exactly  where the money is. is helping athletes train, [11:35] is helping referees judge, is engaging our fans or  making them lose attention if we don't do things [11:41] right. So before we advance any further,  we have to ask ourselves, can we trust AI? [11:48] Our answer is the trustworthy AI framework  that we developed with our partner Deoid. [11:54] It has three components. The first component is  training because not everybody understands what [11:59] is safe and secure or what is role-based security.  We need to break up the level with training. [12:05] The second component, what you're seeing  on your screen are the guidelines. [12:10] Seven pillars of trust. If you make sure your  AI solution has these seven pillars, you will [12:15] be fine. You're not crossing any line. And the  third element is the most practical, a checklist. [12:23] In today's world of AI, everybody's expecting  complicated agents, impossible dashboards. [12:30] But sometimes the solution is a Excel. you can  print and with a pen you can actually check. [12:35] Imagine you're a small national Olympic  committee or international federation and a [12:39] big technological company knocks on your door with  an offer you cannot reject. It's innovative. It's [12:45] cool. The financials look great. Let's do it. But  you don't really understand it. What can you do? [12:51] You have two risks. Risk number one, you reject  because you don't want to send what you don't [12:57] understand. So you missed an opportunity.  others will advance and you will stagnate. [13:02] Risk number two, you sign because you don't want  that, but you cross a line you didn't want crossed [13:07] because you didn't understand what you were  signing. This checklist and the trustworthy [13:11] AI framework help you with these risks. You can  get the checklist and have a conversation with [13:16] this technological partner. Have you thought of  data privacy? Where are you storing the data? [13:21] If you go through all of these and address the red  flags or the not satisfactory answers again with [13:27] the guidelines to find out what's going on, then  you can be sure that you're innovating safely. [13:35] Let's take now a minute to look back at our AI  journey and reflect what what were our challenges, [13:42] lessons learned, and the promise blueprint  for AI transformation. First the challenges [13:49] at the IOC we have a lot of data. I'm sure you  have also a lot of data in your organizations. [13:55] However, I'm also sure that your data is not  always AI ready. Maybe it's not clear. Maybe it's [14:01] not structured. Maybe it's obsolete. You cannot  scale. That was the first challenge we encounter. [14:07] The second is regulatory complexity. We are  literally everywhere. Each nation or region has [14:14] particular regulations regarding AI which are not  static but they change quickly. Staying up to date [14:22] for everywhere in the world is a challenge.  Security risks. It is obvious that a director [14:29] of HR should have different access to information  in your company as a trainee who joined last week. [14:35] However, if an agent has access to the whole  information and both are using the same agent, how [14:42] do you make sure that the agent doesn't answer to  the trainee, what is the salary of the director? [14:48] And the fourth challenge we found is a fragmented  ecosystem. It's difficult to advance fast if you [14:54] have different stakeholders with very different  levels of AI literacy, ambition, even willingness [15:00] to implement AI. Now if you turn the Olympic medal  behind every challenge there is a lessons learned. [15:08] Data foundations first. You need to sharpen the  axe before chopping any trees. Every hour you [15:15] invest in structuring, cleaning, preparing your  data will save you many hours down the road. [15:22] Governance by design. I believe every  company thinks of governance. Some of [15:27] them at the beginning and some of them after the  crisis. You want to belong to those that think of [15:33] governance at the very beginning. There is  no other way. Security at scale. What I was [15:38] mentioning about the trainee and the director,  you need to embed security access based on roles. [15:44] Depending on your role, the agent will answer you  different information. You cannot have it all. [15:49] And collaboration is key. As I was saying, we have  a fragmented ecosystem. It's a difficult scenario. [15:56] We will probably never have the fastest  solution. However, when we find a solution, [16:01] I believe it can be the best solution because we  put in the table those that don't believe in AI [16:06] and those that are enthusiasts, those that know  nothing about and those that are already experts. [16:10] If we have a solution that works for everybody,  this is going to be a hell of a solution. [16:17] Now, the promise blueprint for AI transformation, [16:21] how did we manage to transform our  organization and start implementing AI? [16:26] And I want you to think you are entering  your car and you want to drive it. [16:31] First you need to know where do you want to go.  This is what we call strategic focus for us is [16:37] the five focus areas and our ambition. That's what  we want to achieve. That's where we want to drive. [16:45] Second, deliver tangible value. You know where  you want to go with your car. Then start driving. [16:50] Collect miles else you're not moving. The best  strategy in the world is not worth the paper [16:56] it uses if it never gets implemented. You need  to start and you need to deliver tangible value [17:01] with your initiatives. This does not include just  collecting proof of concepts that don't bring you [17:07] anywhere. You need to know where you're going  and work in that direction delivering value. [17:12] And once you have started moving, you know  where you're going. Please don't crush the car. [17:18] You need trust. Our answer for that is the  trustworthy AI framework. You may have a different [17:23] one, but you need to set clear guard rates and  governance and make sure keep reviewing that. [17:28] Keep reviewing them so you don't cross a line.  Thank you very much. That's all I have today.