https://openai.com/index/scaling-trusted-access-for-cyber-defense
OpenAI has unveiled GPT-5.4-Cyber, a variant of its latest flagship model GPT-5.4, specifically optimised for defensive cybersecurity use cases, arriving days after rival Anthropic released its own frontier security model, Mythos. Alongside the announcement, OpenAI said it is scaling up its Trusted Access for Cyber (TAC) program to thousands of authenticated individual defenders and hundreds of teams responsible for securing critical software. The move signals a broader industry push to place frontier AI capabilities directly in the hands of defenders, as threat actors increasingly leverage the same underlying technologies for offensive purposes.
OpenAI acknowledged the inherently dual-use nature of AI systems, noting that adversaries could potentially repurpose models fine-tuned for software defence to instead detect and exploit vulnerabilities in widely-used software before patches can be applied. To address this, the company said its goal is to democratise access while minimising misuse through a deliberate, iterative rollout, broadening access for legitimate defenders while continuously strengthening safeguards against jailbreaks and adversarial prompt injections as model capabilities advance. OpenAI also revealed that its AI-powered application security agent, Codex Security, has already contributed to resolving more than 3,000 critical and high-severity vulnerabilities.