OpenAI quietly rolled out a new cybersecurity model that matches Anthropic's restricted Mythos model in capability — raising questions about what actually triggered the Trump administration's export controls on Anthropic. OpenAI's updated GPT-5.5-Cyber achieved eighty-five point six percent on CyberGym, an internal benchmark that tests if an AI can find known software vulnerabilities. Anthropic's Mythos 5 scored eighty-three point eight percent on the same test. Yet OpenAI faced no political pushback, while Anthropic is stuck fighting export restrictions that bar foreign nationals from using its models. The White House and OpenAI declined to comment. OpenAI has meanwhile expanded partnerships to Australia, Canada, France, Germany, Japan, Poland, South Korea, and the EU — but access to Mythos was a major negotiation point at last week's G7 Summit. According to Axios, reports suggest personality clashes between Anthropic and the Trump administration may have played as much a role as safety concerns in prompting the export directive.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton