OpenAI rolls out GPT‑5.4‑Cyber

OpenAI has started a limited rollout of GPT-5.4-Cyber, a version of its latest model tuned for defensive cybersecurity work. (usnews.com) The company said on April 14 that access will go first to vetted security vendors, organizations, and researchers because the model is more permissive on sensitive cyber tasks such as vulnerability research and analysis. (usnews.com) OpenAI is also expanding its Trusted Access for Cyber program, launched in February, to thousands of verified individual defenders and hundreds of teams that protect critical software. New verification tiers will determine who can use stronger capabilities, including GPT-5.4-Cyber. (tech.yahoo.com) Cybersecurity work here means using an artificial intelligence system to help find, test, and explain software flaws before attackers exploit them. OpenAI’s own preparedness policy lists cybersecurity alongside biological, chemical, and self-improvement risks as one of the frontier capability areas it tracks for severe-harm risk. (openai.com) The gated rollout comes a week after Anthropic announced Claude Mythos on April 7 under a controlled program called Project Glasswing. Reuters reported that Anthropic said the model had already found “thousands” of major vulnerabilities in operating systems, web browsers, and other software. (tech.yahoo.com) OpenAI has been reshuffling its broader model lineup at the same time. Its help center says GPT-5.1 Instant, GPT-5.1 Thinking, and GPT-5.1 Pro were removed from ChatGPT on March 11, 2026, and old conversations were migrated to GPT-5.3 Instant, GPT-5.4 Thinking, or GPT-5.4 Pro. (help.openai.com) ChatGPT’s main consumer setup now centers on GPT-5.3 Instant as the default for logged-in users, with GPT-5.4 Thinking positioned for harder tasks and GPT-5.4 Pro for the longest and most complex workflows. OpenAI says GPT-5.4 Thinking is stronger than earlier versions at coding, research, spreadsheets, and tool use. (help.openai.com) The immediate change is narrow: GPT-5.4-Cyber is not a general public launch, but a controlled test of how far OpenAI will let trusted defenders use a more capable model on real security problems. (usnews.com)