OpenAI Releases GPT-6 Cyber With ‘Critical’ Security Risk Rating

The significance here is less the model itself than OpenAI's own decision to label it critical while still shipping it to a limited, vetted cohort — an explicit bet that controlled defensive access outweighs the risk of proliferation. That trade-off lands squarely on security teams and vulnerability researchers, who gain a powerful auditing tool but now operate in a landscape where comparable capability is presumed to be diffusing. The open question is whether the vetted-access model holds as pressure grows to widen distribution, and how quickly the defensive and offensive use cases prove separable in practice.