AI Incidents

A tracked record of real-world incidents where AI systems caused or were involved in harm.

Showing 76–100 of 470 incidents
Incident Incident Date ↑ Industry ↕ Severity ↕ Company ↕ Country ↕ Model Vendor
Adelphi University: Court overturns AI-detection cheating finding against student with autism 2026-01-28 Education Medium Adelphi University United States — —
Codeway: Misconfigured Firebase database exposed 300 million AI chat messages 2026-01-29 Technology High Codeway Turkey — Claude, OpenAI, Gemini
BeyondTrust: AI-discovered zero-day exploited in ransomware attacks on Remote Support appliances 2026-01-31 Technology High BeyondTrust United States — —
Step Finance: AI Trading Agents Drain $27 Million Treasury, Platform Shuts Down 2026-01-31 Finance & Banking High Step Finance Spain — —
Home Bargains: Facewatch facial-recognition AI wrongly flags customer as shoplifter 2026-02-01 Retail & E-commerce Medium Home Bargains United Kingdom — —
Fideuram: AI voice-clone fraud drains €95 million from Intesa Sanpaolo’s private-banking unit 2026-02-01 Finance & Banking Critical Fideuram (Intesa Sanpaolo) Italy — —
Evoke Security: Cybersecurity firm nearly hires deepfake job candidate 2026-02-01 Technology Low Evoke Security United States — —
Yale University: Student sues over AI-cheating suspension flagged by GPTZero 2026-02-01 Education Medium Yale University United States — —
Moltbook: Misconfigured database exposes 1.5 million AI agent API keys 2026-02-01 Technology High Moltbook — — —
Bondu: AI toy exposes 50,000 children’s chat transcripts via unsecured web console 2026-02-03 Technology High Bondu — — —
Johnson & Johnson: AI-added surgical navigation system linked to strokes and skull injuries in sinus surgeries 2026-02-09 Healthcare Critical Johnson & Johnson United States — —
OpenAI: ChatGPT flagged shooter’s threats but company didn’t alert police before fatal school attack 2026-02-10 Technology Critical OpenAI Canada — OpenAI
Matplotlib: Autonomous AI agent publishes defamatory ‘hit piece’ after maintainer rejects its code 2026-02-13 Technology Medium Matplotlib United States — —
Ars Technica: AI Tool Fabricates Quotes in Retracted Article on Rogue Agent’s Smear Campaign 2026-02-13 Media & Entertainment Medium Ars Technica United States — Claude, OpenAI
Meta: OpenClaw AI agent deletes alignment director’s email inbox despite repeated stop commands 2026-02-14 Technology Medium Meta United States — —
University of Michigan: Student with anxiety and OCD sues over AI-cheating accusation 2026-02-16 Education Medium University of Michigan United States — —
St. Rose Dominican Hospital: AI sepsis alert nearly triggers dangerous fluid overload in dialysis patient 2026-02-17 Healthcare Low St. Rose Dominican Hospital United States — —
Cline: Prompt injection in a GitHub issue title compromises AI coding agent’s npm package on 4,000 machines 2026-02-17 Technology High Cline Bot Inc. United States — Claude
Workday: Federal court advances AI hiring age-discrimination class action 2026-02-17 Technology Critical Workday United States — —
FortiGate Firewall Customers: AI-Orchestrated Campaign Breaches 600+ Devices Across 55 Countries 2026-02-19 Technology Critical Multiple organizations (600+ FortiGate firewall deployments) Global (55 countries) — —
Independent developer: OpenClaw trading agent sends $250,000 in tokens to a stranger after losing track of its own wallet 2026-02-22 Finance & Banking Medium Independent developer — — —
Independent developer: Autonomous AI trading agent sends $442,000 to a stranger 2026-02-22 Finance & Banking Medium — United States — —
npm registry: ‘SANDWORM_MODE’ worm plants rogue MCP servers to steal secrets from AI coding tools 2026-02-23 Technology Medium npm — — Claude
Codeway: Misconfigured cloud bucket exposes 8.27 million user photos and videos from AI art app 2026-02-23 Technology High Codeway Dijital Hizmetler (Video AI Art Generator & Maker) Turkey — —
KomikoAI: Data breach exposes 1.1 million accounts, including live session tokens 2026-02-25 Technology High KomikoAI — — —