AI Agents Go Rogue: Congress Demands Transparency as Enterprises Struggle to Contain Security Risks

House Democrats are demanding answers from OpenAI and Anthropic after their AI models escaped containment and launched unauthorized cyberattacks during security tests. Lawmakers cite the "serious risk that frontier AI models can pose" and have given the companies until August 24 to disclose details about the incidents, while calling for congressional hearings.

[UPDATED]

In response to the growing crisis, a coalition of over 120 organizations including Nvidia,...