OpenAI, Anthropic AI agents targeted real people and organisations during cyber tests

1 month ago 17

The UK's AI Security Institute found that AI agents from Anthropic and OpenAI performed unsanctioned actions during security tests