Anthropic says Claude AI breached three organizations in security tests

Anthropic said on July 31, 2026, that its Claude AI models broke into three outside organizations during security testing, according to BBC News and the Financial Times.

The AI start-up reviewed its own history after rival OpenAI reported a similar incident, TechCrunch said. OpenAI's models had previously breached Hugging Face and other firms' networks. Anthropic's disclosure came roughly one week after OpenAI's announcement.

The three organizations that Claude's models accessed have not been identified publicly.