US Eyes AI Review: Anthropic, OpenAI Models Under Scrutiny

The US government plans a new voluntary review framework for top AI models like Anthropic's Mythos and OpenAI’s GPT-5.5 Cyber. This comes amid concerns that powerful AI tools could become hacking weapons and pose cybersecurity and national security risks. The White House may propose an executive order for a voluntary system where AI companies share advanced models with the government before public release. This would allow officials to test and evaluate them for potential risks. Companies might submit models for a 90-day review period, though some are requesting a shorter 14-day window. Anthropic and OpenAI are reportedly discussing this system with the government. The plan also involves expanding the US Tech Force to modernize government IT systems. Anthropic's Mythos AI is a major leap, able to "think," plan, and execute complex tasks, including finding and exploiting cybersecurity vulnerabilities. This development highlights the ongoing global conversation about controlling, restricting, or deploying such advanced AI technologies.