OpenAI enters the AI cybersecurity race OpenAI has entered the AI cybersecurity race with the launch of its new model, GPT-5.5-Cyber, a move aimed at competing with the Mythos model developed by Anthropic, amid rising global concerns... In this article: OpenAI has entered the AI cybersecurity race with the launch of its new model, GPT-5.5-Cyber, a move aimed at competing with the Mythos model developed by Anthropic, amid rising global concerns over the use of AI models in vulnerability discovery and cyberattacks. The launch of the new model comes at a time when the cybersecurity sector is undergoing a rapid transformation driven by the significant advancement in the capabilities of language models, which can now analyze software systems, discover complex security vulnerabilities, and execute advanced penetration tests faster and more accurately than traditional tools. Direct competition with Mythos OpenAI's move came weeks after the buzz generated by Anthropic's Claude Mythos model, developed under the Glasswing project, where tests showed the model's ability to discover complex security vulnerabilities in browsers, operating systems, and legacy software infrastructure. Reports indicated that the Mythos model sparked widespread concern within the banking and cybersecurity sectors due to its ability to analyze attack chains and link vulnerabilities together in a more sophisticated way compared to previous tools, prompting major tech companies to accelerate the development of similar models. In response, OpenAI launched the GPT-5.5-Cyber model under the Trusted Access for Cyber program, granting access to pre-verified security teams and institutions in an attempt to expand the use of AI-powered cyber defense tools without opening the door to misuse. Advanced offensive and defensive capabilities The new model can analyze code, discover security vulnerabilities, and create attack scenarios to simulate cyber intrusions, in addition to helping test protection systems and verify the effectiveness of security patches within an