After dissing Anthropic for limiting Mythos, OpenAI restricts access to Cyber, too

OpenAI Follows Anthropic’s Lead in Limiting Cybersecurity Tool Releases

Despite previously criticizing Anthropic for restricting its Mythos cybersecurity tool to select users, OpenAI has decided to take a similar approach with its own product, Cyber. CEO Sam Altman announced that OpenAI will initially release GPT-5.5 Cyber only to "critical cyber defenders." Interested users must apply on OpenAI’s website, providing credentials and outlining their intended usage to gain access.

Why Is Access Being Restricted?

Cyber is designed to help organizations conduct penetration testing, spot vulnerabilities, and analyze malware, making it highly valuable for defense but also potentially dangerous if misused. OpenAI’s application process aims to ensure the tool is only used by qualified security professionals for legitimate purposes.

Trusted Access for Cyber Program

OpenAI says it is working with the U.S. government to expand Cyber’s availability and vet users more effectively through its Trusted Access for Cyber (TAC) system. According to a company spokesperson, thousands of vetted cybersecurity experts and teams can now access these advanced tools with fewer restrictions. Access is provided in tiers, allowing recognized critical defenders to use the most capable versions, such as GPT 5.5-Cyber.

Ironically, OpenAI’s approach closely resembles Anthropic’s earlier, criticized methods, highlighting growing industry concerns about the responsible release of powerful AI-driven cybersecurity tools.

This article is a recap of the original story from TechCrunch.