OpenAI launches GPT‑5.4‑Cyber

OpenAI has started offering GPT‑5.4‑Cyber, a restricted model built for defensive cybersecurity work, to vetted security teams and individual defenders. (openai.com) The rollout expands OpenAI’s Trusted Access for Cyber program to “thousands” of verified individual defenders and “hundreds” of teams that protect critical software, the company said on April 14. OpenAI said customers in its highest access tiers can request the model for advanced defensive workflows. (openai.com) Cybersecurity work often starts with software that arrives only as machine code, not readable source code. OpenAI said GPT‑5.4‑Cyber is tuned for tasks such as reverse engineering binaries, analyzing vulnerabilities, and inspecting malware even when source code is unavailable. (openai.com) OpenAI is not releasing this model broadly. The company said the system is “cyber-permissive,” meaning it is tuned to help legitimate defenders do work that ordinary public models more tightly restrict, while access is gated through identity checks and trust reviews. (openai.com) The company framed the move as preparation for more capable models in the coming months. In its March 5 system card for GPT‑5.4 Thinking, OpenAI said GPT‑5.4 was its first general-purpose model with mitigations for “High” cybersecurity capability. (openai.com) The push comes as artificial intelligence companies race to build tools for software defense as well as coding. Reuters reported on April 14 that OpenAI’s launch followed Anthropic’s announcement of its own frontier cybersecurity model, adding competitive pressure around restricted cyber systems. (reuters.com) OpenAI said it has also given GPT‑5.4‑Cyber access to the United States Center for AI Standards and Innovation and the United Kingdom AI Security Institute for evaluations of cyber capabilities and safeguards. The company separately said firms including Cisco, CrowdStrike, Goldman Sachs, Nvidia, Oracle, SpecterOps, and Zscaler have joined the broader program. (openai.com) Outside reporting described the model as having fewer capability restrictions than standard OpenAI deployments, but still limited to approved defenders rather than the public. Quartz reported that the expansion is aimed at professionals and teams responsible for securing critical software. (qz.com) OpenAI’s earlier February launch of Trusted Access for Cyber paired the program with $10 million in application programming interface credits for defensive work. This week’s update keeps that gated-access approach, but adds a model designed specifically for the kind of low-level software analysis that security researchers use to find and fix flaws. (openai.com)