OpenAI is widening its Trusted Access for Cyber (TAC) program, moving from a pilot to thousands of vetted defenders and hundreds of teams protecting critical software. The initiative is backed by a $10M Cybersecurity Grant Program and debuts GPT-5.4-Cyber, a fine‑tuned model that relaxes refusal limits for legitimate security work. OpenAI frames the expansion around three pillars—iterative deployment, ecosystem resilience, and automated identity verification—across industries.

Since 2023 the company has funded defenders through a Cybersecurity Grant Program and introduced Codex Security, an automated scanner that monitors codebases and proposes patches. To date Codex Security has helped fix over 3,000 critical and high‑severity vulnerabilities and supports more than 1,000 open‑source projects, illustrating OpenAI’s shift toward embedding AI directly into the security workflow.

At the highest TAC tier, vetted users receive unrestricted access to GPT-5.4-Cyber, enabling advanced tasks such as binary reverse engineering without source code. By lowering refusal boundaries, the model can analyze compiled binaries for malware signatures and hidden flaws, accelerating incident response. OpenAI’s approach effectively hands frontline defenders a more powerful, purpose‑built AI ally for real‑time defense.