AI Incidents

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

Showing 101–125 of 470 incidents
Incident Incident Date ↑ Industry ↕ Severity ↕ Company ↕ Country ↕ Model Vendor
Thames Valley Police: Facial recognition misidentifies Southampton software engineer as burglary suspect 100 miles away 2026-02-25 Government & Public Sector Medium Thames Valley Police United Kingdom — —
Google: ‘Silent’ Gemini API privilege change leaves 2,863 public API keys exposing private AI data 2026-02-25 Technology High Google United States — Gemini
U.S. Central Command: Overreliance on Palantir’s Maven AI system contributed to deadly Iran school strike 2026-02-28 Government & Public Sector Critical U.S. Central Command (CENTCOM) Iran — Palantir
OpenClaw: ‘ClawJacked’ flaw let malicious websites hijack local AI agents via WebSocket 2026-03-01 Technology Medium OpenClaw — — —
ModelScope: Command-injection flaw in MS-Agent framework lets attackers run arbitrary commands 2026-03-02 Technology Medium ModelScope — — —
Google: ‘Gemini Live’ Chrome flaw let malicious extensions hijack camera, mic, and local files 2026-03-02 Technology Low Google United States — Google
OpenAI: Insurer sues over ChatGPT’s unauthorized practice of law 2026-03-04 Finance & Banking High Nippon Life Insurance Company of America United States — OpenAI
Amity Regional High School: AI grading tool misreads instructions, wrongly penalizes AP Psychology student 2026-03-05 Education Low Amity Regional High School United States — —
DataTalks.Club: Claude Code runs terraform destroy, wipes 2.5 years of production data 2026-03-05 Education High DataTalks.Club Germany — Claude
Amazon: AI coding agent triggers cascading order-system outages 2026-03-05 Retail & E-commerce High Amazon United States — —
DataTalks.Club: AI coding assistant runs destructive Terraform commands, erasing 2.5 years of data 2026-03-07 Education Medium DataTalks.Club — — Claude
UnitedHealth: Federal judge orders disclosure of AI algorithm used to deny Medicare Advantage claims 2026-03-09 Healthcare High UnitedHealth Group United States — —
Amazon, Perplexity: Federal court blocks Comet AI shopping agent over unauthorized account access 2026-03-10 Technology High Amazon, Perplexity United States — Perplexity
Cal AI: Alleged breach exposes data of 3 million+ calorie-tracking app users 2026-03-11 Technology High Cal AI — — —
Six unnamed hedge funds: Autonomous AI trading agents trigger $500 million flash crash 2026-03-11 Finance & Banking Critical Six unnamed hedge funds United States — —
McKinsey: Red-team AI agent breaches internal Lilli platform in two hours 2026-03-11 Consulting & Professional Services Low McKinsey & Company United States — —
xAI: Grok chatbot confirms fabricated Iran war videos as authentic 2026-03-11 Media & Entertainment High xAI United States — Grok
Multiple unnamed enterprises: Hive0163 ransomware group deploys AI-generated Slopoly malware 2026-03-13 Other Medium Multiple unnamed enterprises — — —
City of Athens, Tennessee: Sixth Circuit sanctions attorneys $15,000 each over AI-fabricated citations 2026-03-13 Government & Public Sector Medium City of Athens, Tennessee United States — —
nginx-ui: Unauthenticated MCP endpoint flaw actively exploited for full server takeover 2026-03-15 Technology High nginx-ui — — —
xAI: Tennessee teenagers sue over Grok-powered app generating child sexual abuse material 2026-03-16 Technology High xAI United States Grok xAI
São Paulo Municipal Government: ‘Smart Sampa’ facial recognition system wrongfully detains thousands 2026-03-17 Government & Public Sector High São Paulo Municipal Government Brazil — —
Sears Home Services: Unsecured databases exposed 3.7 million AI chatbot logs and call recordings 2026-03-17 Retail & E-commerce High Sears Home Services United States — —
Koi Security, Palo Alto Networks: AI-hallucinated threat report wrongly links videoconferencing startup to Chinese espionage network 2026-03-18 Technology High Koi Security, Palo Alto Networks United States — —
Meta: AI agent’s unauthorized forum post exposes internal data for two hours 2026-03-18 Technology Medium Meta United States — —