OpenAI: Hey, We Also Have a New Tool That Is So Scarily Powerful We Can’t Release It

OpenAI is reportedly putting the finishing touches on a new cybersecurity service described as so powerful that it cannot be released to the general public. According to Axios, the tool will be offered to a select set of partners to help bolster their defenses. This announcement comes just days after rival Anthropic began rolling out its own 'dangerous' security model, Mythos, positioning OpenAI's move as a competitive response to ensure it isn't left behind in the hype cycle. The new offering is a distinct product, not a new model, and is unrelated to OpenAI's planned 'Spud' release.

OpenAI already operates a 'Trusted Access for Cyber' pilot program, an invitation-only initiative launched after GPT-5.3-Codex that provides partners with more capable models for defensive cybersecurity work. However, the company had not previously marketed this capability as a world-changing product. The new service appears to be an effort to match Anthropic's bold claims, such as Mythos spotting flaws that evaded human detection for decades—claims other researchers have questioned. Beyond cybersecurity, OpenAI is making ambitious financial projections, forecasting $102 billion in advertising sales by 2030, a stark contrast to its projected $2.5 billion in ad revenue for this year.