GPT-5.4-Cyber and scaled Trusted Access for Cyber
GPT-5.4-Cyber, a cyber-permissive GPT-5.4 variant, anchors a Trusted Access for Cyber program scaled to thousands of verified defenders.
Fine-tuned to refuse less on defensive security work, available through identity verification and an extra application for the most permissive tier. Explicitly framed as preparation for more capable models; Simon Willison read it as OpenAI's answer to Anthropic's Claude Mythos restriction.
- Date
- Tuesday, 14 April 2026
- Lab
- OpenAI
- Kind
- model
- Access
- research preview
GPT-5.5-Cyber followed on 2026-05-07 and GPT-5.6-Cyber with Daybreak in 2026-08.
Sources
- openai.com/index/scaling-trusted-access-for-cyber-defense/
- simonwillison.net/2026/Apr/14/trusted-access-openai/
This record was checked against its sources on 6 October 2026. How we check