Recurring AI Incidents
We group documented incidents by what went wrong and which industry it hit. A pattern is listed only when it happened 3+ times, to 2+ different organizations, caused real harm, and has a known defense.

Know the risks around your AI
What kinds of AI are affected and how they fail.
What is affected
AI system types
What goes wrong
Attack & failure methods
Share of the 2,298 incidents in this loaded archive, not the chance of an attack. Records include failures, allegations and disclosed vulnerabilities—not only confirmed exploits. System types use recorded technology labels; methods use rule-based evidence matching, with one primary category per record. Other or unclear records are not counted in these categories. Higher counts can reflect reporting coverage, not greater vulnerability.
Recurring patterns worth watching
Computed from 2,298 documented incidents. 6 patterns pass the bar. Rising/falling speed is not shown yet: too few recent dated records.
- PATTERN #1 · SCORE 117
Private Data Leaks Through AI Tools (Cross-Industry)
IN PLAIN WORDS:General · 17 incidents across 10 organizations · Category: Sensitive information disclosureMITRE ATLAS AML.T0057 · LLM Data Leakage17Incidents10Organizations hit5.8/10Average riskTier 3Money, data or lawsuitsReal cases behind this pattern
- AI chat app exposes 300 million private messages
Source: AIAAIC Repository · adapted by BlacklistAI · CC BY-SA 4.0 · Original report · Report an error
- Anthropic AI agent hacks third-party website with same target name
Source: AIAAIC Repository · adapted by BlacklistAI · CC BY-SA 4.0 · Original report · Report an error
- OpenAI hacked Medicare portal, Prime Minister Anthony Albanese says
Source: AI Incident Database · Original report · Report an error
- Spanish data watchdog publicises first AI agent-linked data breach report
Source: AI Incident Database · Original report · Report an error
- First Agentic AI Data Breach Reported to Spanish Regulator
Source: AI Incident Database · Original report · Report an error
Step-by-step defense plan for this pattern: Coming soon. Run a free assessment - AI chat app exposes 300 million private messages
- PATTERN #2 · SCORE 92
Private Data Leaks Through AI Tools in Technology
IN PLAIN WORDS:Technology · 11 incidents across 8 organizations · Category: Sensitive information disclosureMITRE ATLAS AML.T0057 · LLM Data Leakage11Incidents8Organizations hit5.5/10Average riskTier 3Money, data or lawsuitsReal cases behind this pattern
- OpenAI models escape test, hack Hugging Face
Source: AIAAIC Repository · adapted by BlacklistAI · CC BY-SA 4.0 · Original report · Report an error
- Meta AI agent leaks sensitive company and user data
Source: AIAAIC Repository · adapted by BlacklistAI · CC BY-SA 4.0 · Original report · Report an error
- AI website generation tool Lovable accused of being highly insecure
Source: AIAAIC Repository · adapted by BlacklistAI · CC BY-SA 4.0 · Original report · Report an error
- Google Search indexes Bard personal chats
Source: AIAAIC Repository · adapted by BlacklistAI · CC BY-SA 4.0 · Original report · Report an error
- ChatGPT leaks user conversations, personal information
Source: AIAAIC Repository · adapted by BlacklistAI · CC BY-SA 4.0 · Original report · Report an error
Step-by-step defense plan for this pattern: Coming soon. Run a free assessment - OpenAI models escape test, hack Hugging Face
- PATTERN #3 · SCORE 77
Private Data Leaks Through AI Tools in Business
IN PLAIN WORDS:Business · 8 incidents across 7 organizations · Category: Sensitive information disclosureMITRE ATLAS AML.T0057 · LLM Data Leakage8Incidents7Organizations hit5.9/10Average riskTier 3Money, data or lawsuitsReal cases behind this pattern
- WebinarTV secretly records and turns meetings into podcasts
Source: AIAAIC Repository · adapted by BlacklistAI · CC BY-SA 4.0 · Original report · Report an error
- AI agent hacks McKinsey employee chatbot
Source: AIAAIC Repository · adapted by BlacklistAI · CC BY-SA 4.0 · Original report · Report an error
- Otter AI transcription leaks confidential investor call
Source: AIAAIC Repository · adapted by BlacklistAI · CC BY-SA 4.0 · Original report · Report an error
- North Korea uses AI to covertly place IT workers in western tech companies
Source: AIAAIC Repository · adapted by BlacklistAI · CC BY-SA 4.0 · Original report · Report an error
- Adobe Creative Cloud accused of using customer to train AI systems
Source: AIAAIC Repository · adapted by BlacklistAI · CC BY-SA 4.0 · Original report · Report an error
Step-by-step defense plan for this pattern: Coming soon. Run a free assessment - WebinarTV secretly records and turns meetings into podcasts
- PATTERN #4 · SCORE 65
Private Data Leaks Through AI Tools in Health
IN PLAIN WORDS:Health · 4 incidents across 3 organizations · Category: Sensitive information disclosureMITRE ATLAS AML.T0057 · LLM Data Leakage4Incidents3Organizations hit5.5/10Average riskTier 3Money, data or lawsuitsReal cases behind this pattern
- Otter AI bot leaks hospital patient health information
Source: AIAAIC Repository · adapted by BlacklistAI · CC BY-SA 4.0 · Original report · Report an error
- Perth doctors warned for using ChatGPT to write patient medical records
Source: AIAAIC Repository · adapted by BlacklistAI · CC BY-SA 4.0 · Original report · Report an error
- Perth doctors warned for using ChatGPT to write patient medical records
Source: AIAAIC Repository · adapted by BlacklistAI · CC BY-SA 4.0 · Original report · Report an error
- Crisis Text Line shares users' mental health data with AI company
Source: AIAAIC Repository · adapted by BlacklistAI · CC BY-SA 4.0 · Original report · Report an error
Step-by-step defense plan for this pattern: Coming soon. Run a free assessment - Otter AI bot leaks hospital patient health information
- PATTERN #5 · SCORE 65
Private Data Leaks Through AI Tools in Media
IN PLAIN WORDS:Media · 4 incidents across 3 organizations · Category: Sensitive information disclosureMITRE ATLAS AML.T0057 · LLM Data Leakage4Incidents3Organizations hit6/10Average riskTier 3Money, data or lawsuitsReal cases behind this pattern
- AI-powered spyware used to track Italian journalists
Source: AIAAIC Repository · adapted by BlacklistAI · CC BY-SA 4.0 · Original report · Report an error
- TikTok fined for selling personal data of US users without consent
Source: AIAAIC Repository · adapted by BlacklistAI · CC BY-SA 4.0 · Original report · Report an error
- TikTok UK fined for misusing childrens' data
Source: AIAAIC Repository · adapted by BlacklistAI · CC BY-SA 4.0 · Original report · Report an error
- Greek Intelligence implicated in "Predatorgate" AI-powered spyware scandal
Source: AIAAIC Repository · adapted by BlacklistAI · CC BY-SA 4.0 · Original report · Report an error
Step-by-step defense plan for this pattern: Coming soon. Run a free assessment - AI-powered spyware used to track Italian journalists
- PATTERN #6 · SCORE 64
Private Data Leaks Through AI Tools in Education
IN PLAIN WORDS:Education · 3 incidents across 3 organizations · Category: Sensitive information disclosureMITRE ATLAS AML.T0057 · LLM Data Leakage3Incidents3Organizations hit5.3/10Average riskTier 3Money, data or lawsuitsReal cases behind this pattern
- UBC academic, students accuse Proctorio of privacy abuse
Source: AIAAIC Repository · adapted by BlacklistAI · CC BY-SA 4.0 · Original report · Report an error
- Proctortrack data breach raises privacy concerns
Source: AIAAIC Repository · adapted by BlacklistAI · CC BY-SA 4.0 · Original report · Report an error
- Data breach reveals data of 440,000 ProctorU users
Source: AIAAIC Repository · adapted by BlacklistAI · CC BY-SA 4.0 · Original report · Report an error
Step-by-step defense plan for this pattern: Coming soon. Run a free assessment - UBC academic, students accuse Proctorio of privacy abuse
