OpenAI unveils GPT-5.4-Cyber in limited rollout to counter cyber threats

Technology Apr 15, 2026

OpenAI just unveiled GPT-5.4-Cyber, a new AI model fine-tuned to help defend against cyber threats.
It's rolling out in limited fashion to select security vendors, organizations, and researchers, so not everyone gets access right away.
This move comes after Anthropic's Mythos model, which also targets finding software vulnerabilities.

GPT-5.4-Cyber analyzes software vulnerabilities

GPT-5.4-Cyber stands out because it can actually analyze and research software vulnerabilities—a big deal for anyone working in cybersecurity.
OpenAI is expanding its Trusted Access for Cyber (TAC) program too, letting thousands more verified users tap into the tool at different access levels.
But since this AI is pretty powerful (and potentially risky), only approved users will get the keys for now.