News-OpenAI-Cyber-Restrictions-2026
April 30, 2026 — OpenAI announced it will begin rolling out GPT-5.5 Cyber, its advanced cybersecurity AI tool, to "critical cyber defenders" through a controlled access program — a move that mirrors the same restrictive approach CEO Sam Altman had criticized Anthropic for taking with its Mythos cybersecurity model.[1]
The tool performs penetration testing, vulnerability identification and exploitation, and malware reverse engineering — capabilities that raise concerns about potential misuse. OpenAI requires applicants to submit credentials and describe their planned use through an application process on its website. The company says it is consulting with the U.S. government and working to identify more users with legitimate cybersecurity credentials to expand access.[1]
The controlled rollout is notable given Altman's previous criticism of Anthropic's similar gatekeeping of Mythos, which he had described as "fear-based marketing." Some security researchers had also argued that Anthropic's rhetoric about Mythos's capabilities was overblown. An unauthorized group reportedly managed to gain access to Mythos despite the restrictions, highlighting the challenges of controlling access to powerful AI cybersecurity tools.[1]