OpenAI has announced the release of GPT-5.5-Cyber, a new AI model designed for cybersecurity, which outperforms Claude Opus 4.7 on the CyberGym benchmark. The model is rated as “High” risk internally by OpenAI, but remains below the critical threshold that would trigger strict usage restrictions.

Alongside the release, OpenAI published a plan titled “Cybersecurity in the Intelligence Age,” emphasizing the need for defenders to keep pace with attackers leveraging new technologies. The company is also expanding its TAC program beyond a limited group of cybersecurity researchers to include government agencies, the financial sector, and MSSPs supporting local hospitals, schools, and water supply facilities.

Source: OpenAI Cybersecurity Plan and Sam Altman on X (https://x.com/sama/status/2049712078836170843)