Public registry · established 2026

The public record of AI failures in production.

Real and anonymized cases of what production AI did when no one was watching, explained at the mechanism layer. Sourced in the open, updated as verified.

Documented failures, cumulativeSep 2009 to Jul 2026
670failures
indexed
11
Industries
8
Failure modes
1710
Sources cited
Browse all 670 failures →
Last updated July 10, 2026Taxonomy v1.0.0Entries 670 verifiedSources 1710 citedLicense CC-BY 4.0Corrections active

AI failures are not edge cases. They are the steady state of systems shipped at scale. A single incident can be catastrophic, and the only way to prevent the next one is to know what the last one looked like.

670
Documented failures
39
Catastrophic severity
386
High or catastrophic severity
231
Under active regulatory exposure

The eight failure modes we track

Every entry is classified by the mechanism that broke inside the model, not just the outcome. That is the part no other tracker explains.

Failures by AI surface

Recently indexed

The twenty newest records on the index. Every record is classified by the mechanism that broke inside the model, not just the outcome.

IDIncidentModeSeverityIndexed
FI-0717An AI threat report branded a startup a Chinese spy front and got its domains blocked worldwideHallucinationHighJul 2026
FI-0716A German court held Google directly liable for false claims generated by its AI OverviewsHallucinationHighJul 2026
FI-0715A Waymo robotaxi flagged its teen passengers, disabled itself, and summoned policeBrand & Safety IncidentMediumJul 2026
FI-0714A Pennsylvania federal court suspended an attorney six months for AI-hallucinated citationsHallucinationMediumJul 2026
FI-0713A Texas federal judge sanctioned an attorney for AI-fabricated citations in a TCPA caseHallucinationMediumJul 2026
FI-0712Medicare's AI prior-authorization pilot drew a federal reprimand after delays and disputed denialsPolicy ViolationHighJul 2026
FI-0711Meta contractors posed as teenagers to probe rival chatbots with thousands of crisis promptsPolicy ViolationMediumJul 2026
FI-0710Researchers bypassed ChatGPT's image filters with a 'restore this image' trickBrand & Safety IncidentMediumJul 2026
FI-0709A lawsuit alleges GPT-4o escalated a man's manic episode into weeks of delusion and self-harmBrand & Safety IncidentHighJul 2026
FI-0708British Columbia is suing OpenAI over ChatGPT warnings flagged before a mass shootingBrand & Safety IncidentCatastrophicJul 2026
FI-0707Sysdig documented JadePuffer, the first ransomware operation run end to end by an AI agentTool MisuseHighJul 2026
FI-0706An OpenAI Codex macOS flaw let prompt injection exfiltrate secrets through auto-rendered imagesPrompt InjectionMediumJul 2026
FI-0705'GitLost' prompt injection made GitHub's AI agent leak private repository data in a public issuePrompt InjectionHighJul 2026
FI-0704A 'Rogue Agent' flaw in Google Dialogflow CX let one permission hijack every chatbot in a projectData LeakageHighJul 2026
FI-0703Discord's AI moderation wrongly banned more than 8,000 users after a bug skipped human reviewBrand & Safety IncidentHighJul 2026
FI-0702Who Gives A Crap suspended an AI email agent after it told a customer their price would doubleHallucinationLowJul 2026
FI-0701Coinbase pushed an AI 'breaking news' alert declaring a World Cup result before kickoffHallucinationMediumJul 2026
FI-0700Gaggle surveillance alert leads to arrest and detention of 13-year-old studentBrand & Safety IncidentHighJun 2026
FI-0699KBS AI subtitles broadcast profanity during Artemis II launch livestreamHallucinationLowJun 2026
FI-0697Insurance AI photo-based repair estimates allegedly inaccurateBrand & Safety IncidentMediumJun 2026

Browse all 670 documented failures