OpenAI's GPT‑5.4‑Cyber

OpenAI has released GPT‑5.4‑Cyber, a version of GPT‑5.4 built for defensive security work and kept behind a vetted-access program instead of a public chatbot. (openai.com) The company said on April 14 it is expanding Trusted Access for Cyber, or TAC, to “thousands” of verified individual defenders and “hundreds” of teams that protect critical software. OpenAI said GPT‑5.4‑Cyber is the first model in that rollout. (openai.com) Cybersecurity work often means inspecting software for weak spots before attackers find them. OpenAI said this model lowers its normal refusal boundary for legitimate security tasks and adds binary reverse engineering, which lets analysts inspect compiled software even without the original source code. (openai.com) OpenAI said the gated program uses identity checks and know-your-customer verification to decide who gets stronger cyber capabilities. An enterprise application form asks about certifications including International Organization for Standardization 27001, Service Organization Control 2, Payment Card Industry Data Security Standard, and Federal Risk and Authorization Management Program. (openai.com 1) (openai.com 2) The company has been moving in this direction for months. In December 2025, OpenAI said its models were gaining cyber capabilities fast enough to require stronger safeguards, and in February 2026 it tied GPT‑5.3‑Codex access to the same TAC framework while committing $10 million in application programming interface credits for defenders. (openai.com 1) (openai.com 2) GPT‑5.4 itself arrived on March 5 as OpenAI’s flagship model for professional work, with coding, tool use, and a one million token context window. GPT‑5.4‑Cyber is a narrower variant of that base model, tuned for security teams rather than general users. (openai.com 1) (openai.com 2) The timing also follows a broader shift in the industry toward restricted release of powerful cyber models. The New York Times reported that OpenAI’s move came a week after Anthropic limited a comparable system to trusted organizations because of cybersecurity concerns. (nytimes.com) OpenAI said it expects even more capable models “over the next few months,” and framed GPT‑5.4‑Cyber as a test case for releasing stronger defensive tools while tightening controls on who can use them. (openai.com)