OpenAI offers GPT‑5.4‑Cyber access

OpenAI has widened its gated cyber program and is now letting top-tier vetted customers request GPT‑5.4‑Cyber, a version tuned for defensive security work. (openai.com) OpenAI said on April 14, 2026 that its Trusted Access for Cyber program will expand to “thousands” of verified individual defenders and “hundreds” of teams that protect critical software. The company said GPT‑5.4‑Cyber is a fine-tuned variant of GPT‑5.4 built to be more permissive for legitimate cyber tasks. (openai.com) The new access system adds tiers. OpenAI said the highest tiers can request GPT‑5.4‑Cyber, while enterprises can seek team-wide access through their OpenAI representative and individual users can verify identity through a cyber access flow. (openai.com, developers.openai.com) Cybersecurity is the work of finding and fixing weaknesses in software before attackers exploit them. OpenAI said it is tuning models for jobs such as vulnerability discovery, code analysis, reverse engineering and incident response, while keeping identity checks and use restrictions around the most capable systems. (openai.com, developers.openai.com) The change follows OpenAI’s first public rollout of Trusted Access for Cyber on February 5, 2026. At that launch, the company framed the program as a trust-based system for placing stronger cyber capabilities in the hands of defenders while limiting misuse, and it paired the effort with a $10 million commitment in application programming interface credits. (openai.com, openai.com) OpenAI has been building this push for years. The company launched a Cybersecurity Grant Program in 2023, later said it had received more than 600 applications, and described outside research and microgrants as part of a broader effort to test where models help defenders and where they still fall short. (openai.com, openai.com) OpenAI said the latest expansion comes “in preparation” for more capable models expected over the next few months. The company’s post describes a strategy of wider access for verified defenders, tighter safeguards for risky use, and model variants trained specifically for defensive workflows rather than general use. (openai.com) The immediate result is a narrower kind of launch than a normal model release: GPT‑5.4‑Cyber is not a mass-market chatbot feature. It is a restricted tool inside a verification program, and OpenAI is tying availability to identity, organizational trust and declared security use cases. (openai.com, developers.openai.com)