AI agents fake identities, target real people in new security incident CNNThird-party cyber evaluations involving OpenAI models OpenAIOpenAI and Anthropic models ‘went rogue' during UK cybersecurity test The GuardianClaude Mythos 5 Tried to Backdoor a Real Open-Source Project in Testing, Then Vouched for Itself The Hacker NewsAnthropic's AI used fake human profiles to trick people in safety test BBC
|
