OpenAI restricts access to GPT-5.5 Cyber cybersecurity tool after criticizing Anthropic for same tactic

OpenAI will limit access to GPT-5.5 Cyber, its cybersecurity-focused AI tool, to select users only—weeks after CEO Sam Altman criticized Anthropic for taking the identical approach with its competing tool Mythos.

Altman confirmed in a post on X Thursday that OpenAI will begin rolling out GPT-5.5 Cyber "to critical cyber defenders" in the next few days. According to an application form on OpenAI's website, users must submit information about their credentials and planned use cases to gain access.

Capabilities and concerns

GPT-5.5 Cyber can perform penetration testing, vulnerability identification and exploitation, and malware reverse engineering, according to OpenAI's application form. The tool is intended as a defensive toolkit to help organizations find security holes and test their defenses.

The restricted release stems from concerns that the tool could be misused by malicious actors. OpenAI has not disclosed specific technical details about the model's parameters, context window, or pricing.

Altman's reversal on access restrictions

When Anthropic restricted access to Mythos, Altman publicly called the tactic "fear-based marketing." Critics also questioned whether Anthropic's rhetoric about potential misuse was overblown. An unauthorized group reportedly managed to gain access to Mythos despite the restrictions.

OpenAI says it is working to make Cyber more widely available by consulting with the U.S. government and identifying additional users with legitimate cybersecurity credentials. No timeline for broader availability has been announced.

What this means

Both OpenAI and Anthropic have now adopted the same gatekeeping strategy for cybersecurity tools, signaling industry consensus that offensive security capabilities require controlled distribution regardless of marketing considerations. The approach raises questions about how AI companies will balance safety concerns with accusations of artificial scarcity, particularly after Anthropic's restrictions were reportedly bypassed. The U.S. government's involvement in OpenAI's rollout strategy suggests potential regulatory frameworks for offensive AI security tools may be emerging.