OpenAI Launches GPT-5.4-Cyber to Limited Users

OpenAI began rolling out GPT-5.4-Cyber, a cybersecurity AI model, to participants in its Trusted Access for Cyber program on 12 April 2026.
The model aims to detect and fix software vulnerabilities, offering defensive use cases while maintaining reduced constraints for vetted testers.
Anthropic’s Mythos model, released a week earlier, prompted warnings from Secretary Scott Bessent and Fed Chair Jerome Powell about AI‑driven cyber risks.
OpenAI plans to expand the program from hundreds to thousands of verified defenders in coming weeks, emphasizing democratized access with identity verification.