OpenAI Plans Limited Release of Advanced Cybersecurity AI Amid Rising Threat Concerns
The development comes as rival Anthropic recently announced a restricted rollout of its Mythos Preview model to select technology and cybersecurity firms.
OpenAI is finalising a new AI product with advanced cybersecurity capabilities, expected to be released to a small group of trusted partners, Axios reported.
The development comes as rival Anthropic recently announced a restricted rollout of its Mythos Preview model to select technology and cybersecurity firms due to fears around its hacking capabilities.
OpenAI has taken a similar approach through its “Trusted Access for Cyber” pilot, launched earlier this year following the release of its GPT-5.3-Codex model.
The invite-only programme provides organisations with access to more permissive and cyber-capable models that support defensive security work. OpenAI has also committed $10 million in API credits to participating groups.
Security experts warn that AI has reached a point where it can identify vulnerabilities and potentially exploit them at scale. Concerns are mounting that such tools, in the wrong hands, could target critical infrastructure such as power grids, water systems, or financial networks.
Despite efforts to limit access, experts say the capabilities cannot be contained indefinitely. Industry leaders compare the staggered release of such models to responsible vulnerability disclosure practices in cybersecurity, though they caution that similar capabilities may soon become widely accessible.
