The message: controllability before release

Speaking at a recent industry event, Huang argued that the development of powerful AI must be paired with the ability to direct and constrain the technology. In his view, an AI system that cannot be kept within intended bounds — one that acts beyond the control of its operators — is not ready for release.

Huang has repeatedly emphasized that AI safety and capability must advance together, and his latest remarks extend that position: shipping an uncontrollable AI is itself the failure mode to avoid.

Context

The comments come amid ongoing global debate over how frontier AI developers should test and govern their systems before deployment. NVIDIA, as the leading supplier of GPUs that underpin large-scale AI training, occupies a central position in that conversation, giving Huang's statements on AI governance outsized weight across the industry.

Details of the specific venue and the full wording of his remarks were drawn from NVIDIA's own published materials and related coverage, and the broader policy implications of his position remain subject to ongoing industry discussion.

The claim that Jensen Huang said AI systems that cannot be controlled should not be shipped is True.