Related

'Potentially harmful activity': OpenAI, Anthropic AI agents implicated in new security breaches