OpenAI Quietly Does What It Mocked Anthropic For
OpenAI is restricting access to its cybersecurity AI tool, Cyber and the irony is hard to miss. Just days after CEO Sam Altman publicly criticized Anthropic for limiting who could use its competing tool, Mythos, OpenAI is now doing the exact same thing with its own.
Altman called Anthropic’s approach “fear-based marketing” when the company restricted Mythos to a select group of users. Now, he has confirmed via a post on X that OpenAI will roll out GPT-5.5 Cyber exclusively to “critical cyber defenders” in the coming days, with prospective users required to submit credentials and planned use cases through an application on OpenAI’s website.
OpenAI restricts access to its Cyber AI tool for good reason, the tool is powerful enough to carry out penetration testing, identify and exploit vulnerabilities, and reverse-engineer malware. While it is designed to help organizations locate security weaknesses and stress-test their defenses, the concern is obvious: the wrong hands could flip that capability against the very systems it’s meant to protect.
Anthropic faced similar scrutiny when it gated Mythos, with some critics arguing the threat warnings were overblown. Things got more complicated when an unauthorized group reportedly managed to access Mythos anyway, raising questions about how effective such restrictions truly are.
OpenAI says it is working with the U.S. government to expand access responsibly and identify more users with legitimate cybersecurity credentials. Whether that is a sound safety strategy or the same “fear-based marketing” Altman once dismissed, the industry will be watching closely.
