Key Takeaways
- OpenAI is rolling out GPT-5.5 Cyber to vetted cybersecurity defenders, mirroring the restricted-access model it publicly mocked Anthropic for using with Mythos
- Sam Altman previously called Anthropic’s limited release “fear-based marketing”, comparing it to selling a bomb shelter after threatening to drop a bomb
- Anthropic’s Mythos was accessed by an unauthorised group via a third-party vendor on the same day it launched
- OpenAI’s Trusted Access for Cyber (TAC) programme has scaled to thousands of verified defenders and hundreds of teams protecting critical software
OpenAI CEO Sam Altman spent weeks ridiculing Anthropic for gatekeeping its cybersecurity model Mythos behind an exclusive enterprise rollout. Then OpenAI did the same thing.
we’re starting rollout of GPT-5.5-Cyber, a frontier cybersecurity model, to critical cyber defenders in the next few days.
we will work with the entire ecosystem and the government to figure out trusted access for cyber; we want to rapidly help secure companies/infrastructure.
— Sam Altman (@sama) April 30, 2026
During a podcast appearance in April, Altman called Anthropic’s approach “fear-based marketing.” “It is clearly incredible marketing to say, ‘We have built a bomb, we are about to drop it on your head. We will sell you a bomb shelter for $100 million,'” he said.
There are people in the world who, for a long time, have wanted to keep AI in the hands of a smaller group of people. You can justify that in a lot of different ways. — Sam Altman, CEO, OpenAI
OpenAI’s Trusted Access for Cyber (TAC) programme offers tiered access to GPT-5.4-Cyber and the forthcoming GPT-5.5-Cyber, fine-tuned for penetration testing, vulnerability identification, and malware analysis. The underlying logic is identical to Anthropic’s: cybersecurity AI is inherently dual-use, and restricting it to vetted defenders limits misuse while creating a premium enterprise tier.
Critics have noted both strategies also protect against distillation, where smaller companies replicate frontier models cheaply, keeping enterprise contracts flowing to the labs that can afford to scale.
Anthropic’s controlled rollout meanwhile backfired immediately. An unauthorised group accessed Mythos on launch day by guessing its URL, exposing the limits of trust-based access controls. OpenAI requires Advanced Account Security for its highest-tier TAC members from June 1, 2026, whether that proves more resilient remains to be seen.
