OpenAI unveils GPT‑5.4‑Cyber
OpenAI released GPT‑5.4‑Cyber, a version of its flagship model tuned specifically for defensive cybersecurity and offered to vetted security professionals rather than the general ChatGPT audience. OpenAI says the model targets thousands of individual defenders and hundreds of security teams and follows rapid capability improvements on capture‑the‑flag benchmarks reported earlier this cycle. ( )
OpenAI has released GPT-5.4-Cyber, a version of its flagship model built for defensive cybersecurity and restricted to vetted users. (openai.com) The company said on April 14 that GPT-5.4-Cyber is being rolled out through its Trusted Access for Cyber program to “thousands” of verified individual defenders and “hundreds” of teams that protect critical software. Reuters reported the model will initially go to vetted security vendors, organizations, and researchers. (openai.com) (money.usnews.com) OpenAI said higher verification tiers unlock stronger capabilities, and the highest tier gets access to GPT-5.4-Cyber, which Reuters described as having fewer restrictions on vulnerability research and analysis than general-purpose releases. (money.usnews.com) (openai.com) Cybersecurity models are built to help defenders find flaws before attackers do, including in compiled software that can be inspected without source code. OpenAI said it is fine-tuning models for “cyber-permissive” defensive use as more capable systems approach release in the next few months. (openai.com) (siliconangle.com) The timing follows a week of new competition in restricted cyber models. Reuters said Anthropic announced Claude Mythos on April 7 under a controlled program called Project Glasswing for select organizations doing defensive cybersecurity work. (money.usnews.com) OpenAI’s broader GPT-5.4 model arrived on March 5 in ChatGPT, the application programming interface, and Codex, with up to 1 million tokens of context and new computer-use features for agents that operate software across applications. (openai.com) The company began Trusted Access for Cyber on February 5 as a pilot tied to GPT-5.3-Codex and paired it with $10 million in application programming interface credits for cyber defense work. OpenAI has separately classified GPT-5.3-Codex as having “High Cybersecurity Capability” under its Preparedness Framework, triggering added safeguards in the application programming interface. (openai.com) (developers.openai.com) OpenAI said the new release is part of a trust-based system rather than a public chatbot launch: identity checks, tiered access, and tighter limits on who can use the most permissive tools. The company is betting that narrower distribution can expand help for defenders without opening the same capabilities to everyone. (openai.com)
