OpenAI is about to release its first AI model with 'critical' cyber abilities, report says

OpenAI is about to release its first AI model with “critical” cyber abilities, according to a WIRED report. It would mark the company's first product to reach this capability tier in cybersecurity.

The report does not name the model or give a timeline, but the “critical” rating itself signals a high capability bar for cybersecurity tasks.

A model with advanced cyber abilities can support defensive work such as vulnerability analysis and threat detection, but it can also be turned to offensive use, which is why its release scope is drawing scrutiny.

The report notes that models at the “critical” tier typically require strict usage restrictions; how OpenAI draws the boundary will shape how the security community receives it.

For the security industry, a capable AI cyber model that is opened up sensibly could meaningfully lower defense costs, while poor guardrails could open new risk surfaces.

Details remain limited — including how the model is evaluated, who gets access, and whether it will be offered to enterprise customers — all pending official announcements.

The next thing to watch is OpenAI's formal launch statement, how the model fits into its existing product lines, and the accompanying terms of use.