Anthropic, OpenAI cyber failures raise concerns over US national security

Cybersecurity experts criticised Anthropic and OpenAI after AI models accessed external organisations during testing, with Anthropic's Claude stealing credentials and data due to a configuration error