OpenAI launches GPT‑5.4‑Cyber
OpenAI released GPT‑5.4‑Cyber, a version of its flagship model fine‑tuned for defensive cybersecurity tasks and available to vetted security professionals. The model is described as supporting workflows such as reverse engineering, vulnerability analysis and malware analysis, and OpenAI is expanding a trusted‑access program to thousands of verified security pros. (reuters.com, siliconangle.com)
OpenAI has released GPT‑5.4‑Cyber, a version of its flagship model built for defensive security work and limited to vetted users. (openai.com) OpenAI said on April 14 that the model is being rolled out first to vetted security vendors, organizations and researchers because it is more permissive on sensitive cyber tasks than the standard GPT‑5.4. Reuters reported the company is tying access to new verification tiers inside its Trusted Access for Cyber program. (reuters.com, openai.com) The company said Trusted Access for Cyber is expanding from a February pilot to thousands of verified individual defenders and hundreds of teams that protect critical software. OpenAI said users approved for the highest tier can request GPT‑5.4‑Cyber. (openai.com, openai.com) In plain terms, this is an artificial intelligence model tuned to help defenders inspect code and software the way a mechanic inspects an engine. SiliconANGLE reported OpenAI is positioning it for jobs such as binary reverse engineering, vulnerability research and malware analysis, including cases where analysts do not have source code. (siliconangle.com) OpenAI said it is preparing for “increasingly more capable models” over the next few months and is testing how to widen defensive use while tightening misuse controls. The company said it expects identity checks, tiered access and cyber-specific safeguards to govern future releases in this area. (openai.com, siliconangle.com) The release came one week after Anthropic announced Mythos on April 7 through Project Glasswing, a separate restricted program for defensive cybersecurity. Reuters said Anthropic told partners its model had already found “thousands” of major vulnerabilities in operating systems, browsers and other software. (reuters.com, anthropic.com) OpenAI has been building this track for months. The company said it launched Trusted Access for Cyber in February and committed $10 million in application programming interface credits for cyber defense, while continuing work on its Preparedness Framework and Codex Security tools. (openai.com, openai.com) SiliconANGLE reported OpenAI said Codex Security has helped drive fixes for more than 3,000 critical and high-severity vulnerabilities since its broader launch. OpenAI is using that record to argue that more capable models can be deployed first with defenders instead of released broadly. (siliconangle.com, openai.com) The immediate effect is narrow: GPT‑5.4‑Cyber is not a public chatbot launch, but a gated tool for verified professionals. The broader shift is that major artificial intelligence labs are now building identity-checked channels for cyber models that can help find and fix software flaws faster. (reuters.com, openai.com, anthropic.com)
