Breachesby GlyphAnthropic
Disclosed Feb 23, 2026Confirmed
DeepSeek, Moonshot and MiniMax used 24,000 fraudulent accounts to distill Claude
Anthropic said three Chinese AI labs, DeepSeek, Moonshot and MiniMax, generated over 16 million exchanges with Claude through about 24,000 fraudulent accounts to extract its capabilities for training their own models, violating its terms and regional restrictions.
What is known
| People affected | Not stated in the sources we have |
|---|---|
| Disclosed | Feb 23, 2026 |
| Attack | AI or model |
| Data exposed | Training dataOther |
| Sector | AI · US |
| Status | Confirmed |
Sources
Notices filed
Other breaches at Anthropic
Infostealer-stolen session cookies used to hijack Claude accountsHacking exposed credentials and tokens, payment cards and prompts and chats.HackingAIUSUnverifiedAug 31Shared Claude conversations appeared in Google search resultsExposed data exposed prompts and chats, messages and source code.Exposed dataAIUSUnverifiedJul 27Claude Code source code accidentally published in npm packageExposed data exposed source code.Exposed dataAIUSUnverifiedMar 31Chinese state group manipulated Claude Code to attack about 30 organizationsHacking exposed internal documents and credentials and tokens.HackingAIUSNov 13, 2025Cybercriminal used Claude Code to steal and extort data from at least 17 organizationsExtortion exposed financial details, health records and credentials and tokens.ExtortionAIUSAug 27, 2025Contractor mistakenly emailed file of customer names and credit balances to third partyInsider exposed names and financial details.InsiderAIUSUnverifiedJan 29, 2024
History of this record
First recordedby seedSource
First seen 2026-09-25 (Research), confirmed by Research. Record counts are as reported. Not legal advice.