OpenAI's Cyber Model

OpenAI has started a limited rollout of GPT‑5.4‑Cyber, a version of its latest model tuned for defensive cybersecurity work. (openai.com) Cybersecurity is the work of finding and fixing weaknesses before attackers use them, and OpenAI said this model is meant to help defenders do that faster. The company said GPT‑5.4‑Cyber can support tasks including vulnerability research, reverse engineering, malware analysis, and incident response. (openai.com) OpenAI said on April 14, 2026 that access is being expanded through its Trusted Access for Cyber program to thousands of verified individual defenders and hundreds of teams responsible for protecting critical software. The company said the model is being released to vetted security professionals rather than through ChatGPT for the general public. (openai.com) (cnet.com) The release follows OpenAI’s March 5, 2026 launch of GPT‑5.4, which the company described as its flagship model for professional work with coding, tool use, computer control, and a context window of up to 1 million tokens. OpenAI’s system card for GPT‑5.4 said that model was the first general-purpose system in the series to include mitigations for what it classified as high cybersecurity capability. (openai.com 1) (openai.com 2) OpenAI framed the new release as part of a broader shift toward controlled access for models that can help defenders but could also be misused. In February 2026, the company introduced Trusted Access for Cyber as an identity-based program and committed $10 million in application programming interface credits for cyber defense work. (openai.com) The company said it is preparing for “increasingly more capable models” in the coming months and is pairing wider defensive access with tighter safeguards. OpenAI’s December 2025 cyber policy post said advanced models can improve defense while also creating dual-use risks that require layered controls and outside partnerships. (openai.com 1) (openai.com 2) Outside reports said the move also lands in the middle of a competition among artificial intelligence companies to build security-focused systems for a narrow set of trusted users. Reuters reported OpenAI’s announcement came one week after Anthropic introduced Mythos under its own controlled cybersecurity program. (reuters.com) (tech.yahoo.com) OpenAI said GPT‑5.4‑Cyber is the first step, not the final shape of the program. For now, the company is keeping the model behind verification checks and trusted-access rules while it tests how far a cyber-permissive system can go without being opened to everyone. (openai.com)