OpenAI has updated its cybersecurity model GPT-5.5-Cyber and expanded the scope of how verified security companies can use it, Axios reported on June 22.

OpenAI said in a blog post that the update means GPT-5.5-Cyber "lets approved security experts do more work and has improved its ability to carry out security tasks."

The updated version includes deep analysis of large codebases, identification of security vulnerabilities, vulnerability verification, and software patch development and testing. In the CyberGym benchmark test, which measures whether an AI agent can reproduce known software vulnerabilities in a test environment, GPT-5.5-Cyber scored 85.6 percent, ahead of GPT-5.5 at 81.8 percent.

OpenAI also introduced the OpenAI Daybreak Cyber Partner Program to help verified cybersecurity firms use its models to secure customer environments. The program allows participating security companies to use GPT-5.5 Trusted Access for Cyber in the products and services they provide to their customers.

Previously, approved organisations could mainly use the cyber model only on systems they owned or had permission to test.

OpenAI is also supporting the Patch the Planet initiative in cooperation with Trail of Bits, HackerOne and Calypso to help open-source maintainers handle vulnerabilities found with AI-assisted tools.

OpenAI said it has formed partnerships with agencies in Australia, Canada, France, Germany, Japan, Poland, South Korea and the European Union, and is also working with major infrastructure operators and government networks.