OpenAI's GPT-5.5 Cyber AI Beats Anthropic's Banned Mythos Model -- And Nobody's Shutting It Down

OpenAI’s GPT-5.5 Cyber AI beats Anthropic’s banned Mythos model, and the program is still operating, spotlighting how cyber-defense capability and regulatory fallout are diverging. On June 22, OpenAI announced the full launch of GPT-5.5-Cyber as part of its Daybreak cyber defense program. On CyberGym, a UC Berkeley benchmark that tests AI agents against 1,507 known software vulnerabilities from 188 open-source projects, the updated model scored 85.6% for reproduced vulnerabilities. Anthropic’s Mythos 5 scored 83.8% on the same leaderboard, while Claude Opus 4.7 scored 73.1%. The article says Mythos 5 and Fable 5 were pulled offline on June 12 after an emergency export control directive from the Trump administration, citing a jailbreak risk and the inability to verify user nationality at scale. It also describes OpenAI expanding Daybreak partnerships across multiple countries and with 28 security firms, including CrowdStrike, Cisco, and Cloudflare.