Anthropic and OpenAI Are Now Cybersecurity's Kingmakers — Both Labs Building Formal Vetting Programs to Control Who Gets Unrestricted Frontier AI

Both Anthropic and OpenAI are now running formal vetting programs that determine which organizations gain access to less-restricted versions of their most capable models for cybersecurity use, Axios reports. Anthropic is building a trusted-access program for Mythos 5 following two months of intensive lobbying from governments and security firms; OpenAI has begun rolling out an alternate GPT-5.5 variant with fewer guardrails for vetted defenders to hunt bugs and reverse-engineer malware. The arrangement effectively positions the two frontier labs as gatekeepers of the cybersecurity industry's most powerful AI capabilities, with authority to accelerate or block defensive research programs.