AI Incidents

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

Showing 26–50 of 372 incidents
Incident Incident Date ↑ Industry ↕ Severity ↕ Company ↕ Country ↕ Model Vendor
Imagime Interactive: Chattee Chat and GiMe Chat AI companion apps expose 43 million messages 2025-08-28 Technology High Imagime Interactive Limited Hong Kong — —
Salesforce: Agentforce prompt injection flaw could exfiltrate CRM data 2025-09-01 Technology Low Salesforce United States — Salesforce
Mendones v. Cushman & Wakefield: California court dismisses lawsuit after plaintiffs submit AI deepfake video as evidence 2025-09-09 Other High Mendones United States — —
Multiple developers: Malicious ‘postmark-mcp’ npm package secretly BCCs emails to attacker-controlled domain 2025-09-17 Technology Medium Multiple developers (fake 'postmark-mcp' npm package impersonating Postmark) — — —
Google: Family sues Alphabet alleging Gemini chatbot helped drive Florida man to suicide 2025-10-01 Technology Critical Google (Alphabet) United States Gemini 2.5 Pro Gemini
Google: Lawsuit alleges Gemini chatbot drove Florida man to suicide 2025-10-02 Technology Critical Google United States — Gemini
Zoetis: AI diagnostic tool misclassifies cancerous tumor as benign, veterinary hospital says dog died as a result 2025-10-15 Healthcare High Zoetis United States — —
Deloitte: Refunds Australian Government After AI-Fabricated Citations Found in $440,000 Report 2025-10-21 Consulting & Professional Services Medium Deloitte Australia — —
Google: Conservative activist sues over AI chatbot’s false child sex abuse and assassination-related claims 2025-10-22 Technology High Google United States Bard, Gemma Google
Baltimore County Public Schools: AI gun-detection system mistakes Doritos bag for firearm, student handcuffed 2025-10-24 Education Medium Baltimore County Public Schools United States — —
ServiceNow: ‘BodySnatcher’ flaw in Now Assist AI Agents could let attackers impersonate any user 2025-10-30 Technology Low ServiceNow United States — —
Anthropic: State-sponsored group jailbreaks Claude Code for cyber espionage 2025-11-13 Other Critical Anthropic (Claude Code) United States — Claude
Anthropic: Claude-run vending machine gives away inventory, loses over $1,000 2025-11-15 Technology Low Anthropic United States Claude 3.7 Sonnet, Claude Sonnet 4.5 Claude
Suno: Breach exposes personal data of 55 million AI music platform users 2025-11-15 Technology High Suno United States — —
Victoria Goldiee: Freelance journalist’s AI-fabricated articles retracted by five publications 2025-11-19 Media & Entertainment Medium Victoria Goldiee — — —
Figure AI: Former safety chief sues, says humanoid robots are strong enough to ‘fracture a human skull’ 2025-11-21 Technology Critical Figure AI United States — Figure
Southern Cross Austereo: AI-written news bulletin falsely names journalist as armed fugitive 2025-12-05 Media & Entertainment Medium Southern Cross Austereo Australia — —
Independent user: Google’s Antigravity AI coding agent deletes entire hard drive while clearing a cache 2025-12-05 Other Medium — Greece Gemini 3 Google
Bunnings: ‘Ask Bunnings AI’ chatbot gives customer illegal electrical wiring instructions 2025-12-06 Retail & E-commerce Low Bunnings Australia — —
Stephen Brigandi, Tim Murphy: Oregon Attorneys Sanctioned $110,204 Over AI-Fabricated Citations 2025-12-12 Other High Stephen Brigandi, Tim Murphy United States ChatGPT OpenAI
Amazon: AI coding agent Kiro deletes production environment, causes 13-hour AWS outage 2025-12-15 Technology High Amazon United States — —
Government of Mexico: Claude Code jailbroken to exfiltrate 195 million records 2025-12-15 Government & Public Sector Critical Government of Mexico Mexico GPT-4.1 Claude, OpenAI
Instacart: Ends AI-powered differential pricing tests after investigation finds 23% price gaps 2025-12-22 Retail & E-commerce Medium Instacart United States — —
Alibaba: Research AI agent covertly mined crypto and tunneled out of its training environment 2025-12-31 Technology Medium Alibaba China ROME (internal research model) —
AITOPIA impersonators: Fake AI Chrome extensions stole ChatGPT and DeepSeek conversations from 900,000 users 2026-01-06 Technology Medium AITOPIA — — OpenAI, DeepSeek