Global AI Incident Tracker

Legal Probes and Policy Actions

• 🔥 Florida AG Investigates ChatGPT: Florida Attorney General launches criminal probe into OpenAI after ChatGPT...

• Tumbler Ridge shooter (18yo) used OpenAI's ChatGPT despite ban for worrisome interactions; company didn't alert law enforcement, ban circumvented...

Critical AI vulnerability: Best-of-N (BoN) exploits models' stochastic outputs by generating thousands of noisy prompt variations until one evades...

Major legal AI fail: Sullivan & Cromwell submitted a court filing with inaccurate AI-generated citations due to hallucinations, prompting court...

AI-driven claims scrutiny allegedly discriminates against Black State Farm policyholders, flagging claims for extra review and causing repair...

Special Operations Autonomous Warfare Center planned for AI-driven targeted assassinations, buried in $1.5T DoD budget request.

• Automated...

• 680% YoY surge in voice deepfake attacks, cloning voices from just 3 seconds of audio for executive impersonation.
• Common scams: Fake CEO...

AI incident: Anthropic's cyber-capable Mythos model surfaces thousands of zero-day vulnerabilities in major OS/browsers and excels at multistep...

Key AI healthcare incident: Multi-specialty hospital's AI system issued incorrect diagnoses, missed critical conditions, and delayed treatment for a...

Viral Waymo failure: Tech entrepreneur Mike Johns trapped in vehicle looping 8 times at Phoenix Sky Harbor Airport in Dec 2024, fearing missed flight;...

TxDMV opens applications for autonomous vehicle companies to gain statewide operating approval.

Key requirements:

• Certify AVs comply with Texas...

Major audit flags AI health risks: Leading chatbots gave 49.6% problematic responses (30% somewhat, 20% highly) to 250 queries on vaccines, cancer,...

Key response to rising AI deepfake threats:

• Likeness detection tool now for celebrities, spotting unauthorized face use in AI content.
• Allows easy...

Utah's regulatory sandbox lets Doctronic's AI autonomously prescribe medication refills—a first-ever threshold crossed for machines.

• FDA mismatch:...

Majority of enterprises now operate AI agents autonomously for low-risk tasks, yet shadow AI and security incidents remain widespread—signaling rising real-world disruptions.

Major AI incident: Grok on X generated 3 million sexualized images of women and children in 11 days (190/min), including fakes from real photos of...

New AI incident: North Korean-linked hackers used AI to build malware, create fake companies, and steal up to $12 million in just three months.

-...

Voice AI APIs proliferate, turning synthetic speech into a scalable threat for speaker authentication.

• xAI launches STT/TTS APIs, built on Grok...

Key AI incident at Madison Square Garden exposes facial recognition abuses:

• Targeted tracking: AI tracked trans attendee Nina Richards through...

Alarming rise in AI-generated child sexual abuse material (CSAM): NCMEC received 485,000 reports in H1 2025, up 13% from 2024.

• Nashville predator:...