OpenAI unveils GPT-5.4-Cyber, expands access for verified security experts

SiliconANGLE reports that OpenAI has announced a major expansion of its Trusted Access for Cyber program to thousands of verified security experts, as well as the release of GPT-5.4-Cyber, a refined version of its GPT-5.4 model intended for defensive cybersecurity work.

The new model, which OpenAI refers to as "cyber-permissive," has been specifically designed to lower refusal boundaries for genuine cybersecurity jobs and adds capabilities not found in the ordinary version of GPT-5.4. Binary reverse engineering is one of the new features that enables security experts to examine generated software for viruses, security flaws, and vulnerabilities without requiring access to source code. OpenAI is restricting early access to verified security suppliers, researchers, and organizations due to its permissive nature.

Launched in February with a $10 million grant, OpenAI's Trusted Access for Cyber program provides access through tiered verification levels, with the highest tier unlocking GPT-5.4-Cyber. Businesses can apply for access through their OpenAI representative, and individuals can confirm their identification at chatgpt.com/cyber. Clients who have already signed up for the program may submit separate applications for higher tiers. The announcement coincides with a larger movement that shows how AI developers are addressing cybersecurity risk by shifting from broad capabilities limitations to identity-based access constraints.