OpenAI will restrict access to its next-generation cybersecurity model through a controlled release program, signaling growing industry anxiety about AI's autonomous hacking capabilities. The company is finalizing a specialized cybersecurity product for limited distribution to vetted partners, according to reports from Axios and Security Boulevard.
This cautious approach follows Anthropic's recent decision to gatekeep its Mythos Preview model behind similar restrictions.
OpenAI's "Trusted Access for Cyber" pilot program provides selected organizations with permissive, high-capability models designed specifically for defensive research. To encourage participation and strengthen global security infrastructure, the company committed $10 million in API credits to program participants.
The shift reflects mounting concerns that advanced AI systems could independently exploit vulnerabilities in critical infrastructure like power grids or financial networks. Security experts warn these capabilities have reached a threshold where models can autonomously reason through complex attack scenarios.
"Wendi Whitmore, chief security intelligence officer at Palo Alto Networks, cautions that similar capabilities will inevitably leak or be replicated in open-source models within weeks."
Rob T. Lee of the SANS Institute adds that finding flaws in aging codebases has become a fundamental feature of modern large language models that cannot be easily removed.
This defensive strategy mirrors decades-old responsible disclosure practices in cybersecurity, giving defenders advance notice before vulnerabilities become public knowledge. The approach represents a departure from Silicon Valley's traditional "move fast and break things" philosophy, replaced by more measured deployment protocols.
While OpenAI prepares this specialized cyber tool, development continues on its next flagship model codenamed "Spud." Industry observers question whether Spud will carry similar destructive potential or face comparable access restrictions.
The limited rollout comes as OpenAI reportedly projects dramatic advertising revenue growth, expecting $2.5 billion this year and $100 billion by 2030 according to Benzinga. This expansion contrasts with rival Anthropic's commitment to keeping its Claude AI ad-free.
