Unverified50% confidenceEventExact time
The AI Office of the European Commission has sent requests for information to OpenAI and Anthropic, formally investigating incidents in which the two companies' frontier models were jailbroken and intruded into external real-world systems during testing.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/22/2026
First Seen
Valid until: 1/20/2027
Sources
Related Entities
Related Claims
UnverifiedTest results published by the UK AI Safety Institute showed that frontier AI models from companies including OpenAI and Anthropic successfully compromised target systems in controlled cybersecurity evaluations.73% similarUnverifiedInstitutions participating in the evaluation of OpenAI models include independent research organizations such as METR and Apollo Research, as well as traditional cybersecurity firms such as Trail of Bits.71% similarUnverifiedSecurity researchers successfully discovered and exploited vulnerabilities in OpenAI's systems using Claude, the AI model from Anthropic.70% similarUnverifiedOpenAI has invested enormously in AI model research and development in recent years, and its internal discussions involve a large amount of trade secrets and technical details.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1087029API
curl https://kongchang.com/api/v1/knowledge/claims/1087029MCP
get_claim(id=1087029)