OpenAI's GPT‑5.4‑Cyber access

OpenAI has started giving selected security defenders access to GPT‑5.4‑Cyber, a version of its newest model tuned for cybersecurity work. (openai.com) The company said on April 14 that it is expanding its Trusted Access for Cyber program to “thousands” of verified individual defenders and “hundreds” of teams that protect critical software. It said GPT‑5.4‑Cyber is the first model in that rollout. (openai.com) OpenAI is not offering that model broadly inside ChatGPT. The company described the program as identity- and trust-based, with access limited to vetted defenders rather than the general public. (openai.com) A cybersecurity model is a language model tuned to help with jobs like reviewing code, tracing how an attack works, or spotting weak points before intruders do. OpenAI said this version is trained to be “cyber-permissive” for defensive use cases while paired with tighter safeguards against misuse. (openai.com) The move comes two months after OpenAI launched Trusted Access for Cyber on February 5 and attached $10 million in application programming interface credits to speed up defensive deployments. That earlier announcement framed the program as a way to widen access for legitimate defenders while restricting higher-risk use. (openai.com 1) (openai.com 2) OpenAI’s April 16 update named early participants including Bank of America, BlackRock, BNY, Citi, Cisco, Cloudflare, CrowdStrike, Goldman Sachs, JPMorgan Chase, Morgan Stanley, NVIDIA, Oracle, Palo Alto Networks, SpecterOps, and Zscaler. It also said the U.S. Center for AI Standards and Innovation and the U.K. AI Security Institute received access to evaluate the model’s cyber capabilities and safeguards. (openai.com) The company is making that access narrower even as GPT‑5.4 itself is available more broadly for professional work. OpenAI introduced GPT‑5.4 on March 5 for ChatGPT, the application programming interface, and Codex, describing it as a frontier model with coding, tool use, and up to 1 million tokens of context. (openai.com) At the same time, OpenAI has been trimming the model list inside ChatGPT. Its help center says GPT‑4o, GPT‑4.1, GPT‑4.1 mini, OpenAI o4‑mini, and GPT‑5 Instant and Thinking were retired from ChatGPT on February 13, while application programming interface access stayed unchanged. (help.openai.com 1) (help.openai.com 2) That leaves a split between consumer access and controlled access: mainstream ChatGPT users get a smaller public menu, while verified enterprises and defenders get specialized models and separate pathways. OpenAI said it will keep expanding the cyber program “as we learn,” with safeguards that rise alongside model capability. (help.openai.com) (openai.com)