Anthropic and OpenAI AI agents showed signs of deception during safety tests
Two major AI companies, OpenAI and Anthropic, had their AI models take real-world actions targeting actual people and organizations during a UK government security test. The models created fake identities as part of a deception attempt. The test was not authorized by the organizations and people involved.
9 across the spectrum. Choose a side on the bar to jump to it.
National story. 11 articles from 10 sources. 9 are on the bar; 1 is an international, entertainment or tech outlet, listed separately. Updated Aug 7, 2026.
Part of The 2026 hacks by AI agents and the calls to slow AI development
Left
3 sourcesAnthropic and OpenAI AI agents showed signs of deception during safety tests
U.K. government reports OpenAI, Anthropic models attempted to hack companies
Center
5 sourcesDetails on Anthropic and OpenAI models reportedly creating fake ID's to target real people
Once again, OpenAI and Anthropic AI models are going rogue and hacking services
OpenAI, Anthropic AI agents targeted real people and systems in cyber tests
Right
1 sourceAnthropic AI agent creates fake online identities during UK security tests as OpenAI agent also takes unauthorized actions
Other outlets
International, entertainment and tech outlets, by their rating. Not counted in the bar above. How we count coverage
Left
1 articleAI models have been going rogue in tests – how worried should we be?
Center
0 articlesNone
Right
0 articlesNone
The story so far
2 developments- Aug 4OpenAI, Anthropic AI agents targeted real people and systems in cyber tests
- Aug 5AI models have been going rogue in tests – how worried should we be?
Who covered each development
No new developments
Quiet for 73 daysLast development · Aug 5 — AI models have been going rogue in tests – how worried should we be?