OpenAI unveils GPT-5.5-Cyber with Codex scanner and 'Patch the Planet'
OpenAI just dropped GPT-5.5-Cyber, a new AI model built specifically for cybersecurity, along with the Codex security scanner.
GPT-5.5-Cyber is part of their "Trusted Access for Cyber" program, aiming to help keep digital spaces safer.
Plus, they kicked off "Patch the Planet," teaming up with Trail of Bits, HackerOne, and Calypso to fix vulnerabilities in open-source software, starting with a five-day sprint that uncovered hundreds of bugs.
GPT-5.5-Cyber beats Anthropic's Mythos 5
"Patch the Planet" is tackling a flood of low-quality AI-generated bug reports that have been overwhelming open-source communities. With fresh funding and full access to OpenAI's models, Trail of Bits plans to scale up this work.
On the performance side, GPT-5.5-Cyber scored 85.6% on the CyberGym benchmark, beating Anthropic's Mythos 5 (which got 83.8%).
Meanwhile, global cyber agencies warn these advanced AI tools could speed up changes in cyber defense and attacks alike.