OpenAI has unveiled GPT-5.4-Cyber, a specialized version of its flagship AI model designed for defensive cybersecurity tasks. This announcement follows closely on the heels of rival Anthropic's unveiling of their frontier AI model, Mythos.
Mythos, revealed on April 7, is currently being utilized within Anthropic's "Project Glasswing." This controlled initiative allows select organizations to test the unreleased Claude Mythos Preview model for defensive cybersecurity. The model has reportedly identified "thousands" of significant vulnerabilities across operating systems, web browsers, and other software.
OpenAI, the creator of ChatGPT, stated that GPT-5.4-Cyber will initially be available to a select group of vetted security vendors, organizations, and researchers due to its more permissive architecture. The company is also broadening its "Trusted Access for Cyber" (TAC) program, extending access to thousands of verified individual defenders and hundreds of teams responsible for protecting critical software.
New tiers are being introduced to the TAC program, launched in February. Higher levels of verification will grant access to more advanced capabilities. Users who qualify for the highest tier will be able to utilize GPT-5.4-Cyber, which offers fewer restrictions on sensitive cybersecurity activities like vulnerability research and analysis.