OpenAI has unveiled a new AI model, GPT 5.4 Cyber, specifically designed for cyber defense. This specialized variant of their flagship GPT 5.4 model features fewer restrictions on cybersecurity-related queries for legitimate, defensive applications.
GPT 5.4 Cyber enhances security capabilities with features like binary reverse engineering. This allows researchers to analyze compiled software for malware and vulnerabilities without needing source code access.
Due to its advanced, more permissive nature, OpenAI is implementing a limited rollout. Access is restricted to vetted security vendors, organizations, and researchers through their Trusted Access for Cyber program.
The announcement follows Anthropic's recent release of Claude Mythos Preview. Anthropic stated this model can identify numerous previously unknown, high-severity vulnerabilities but deemed its capabilities too dangerous for a full public release.
Previously, a data leak in late March indicated Anthropic was developing an AI model that its own engineers flagged for "unprecedented cybersecurity risks." This model, Claude Mythos Preview, has been made available in a restricted capacity as part of Anthropic's Project Glasswing, aimed at strengthening critical software.