Jensen Huang calls on AI labs to self-police their models

By Splitfeed EditorialLast updated Sep 28, 2026, 15:10 UTC

Around September 27 to 28, 2026, Nvidia CEO Jensen Huang made a pair of statements that pulled in different directions. On the cautionary side, he called on AI companies, explicitly including OpenAI, to shut down any models they cannot make safe, framing this as an industry responsibility rather than something governments should impose. On the supportive side, he described OpenAI as one of the most consequential companies in history and offered it open use of Nvidia's OpenShell platform, saying OpenAI could use whatever part of it the company wanted.

The pairing matters because it comes from one of the most powerful figures in AI infrastructure. Nvidia supplies the hardware that underlies virtually every major AI training operation, so when Huang speaks about lab conduct, it carries commercial and symbolic weight that an outside critic's remarks would not. His preference for self-policing over government regulation is also a substantive policy position, not just a sentiment, and it arrives at a moment when legislators in multiple jurisdictions are actively debating how much authority to give regulators over foundation models. Whether voluntary self-governance is a credible substitute for binding rules is the question his remarks reopen.

Because no public reactions from the Pro-AI, Anti-AI, or Middle Ground camps had been recorded at the time of writing, what each camp would typically argue can only be anticipated. Pro-AI voices would likely treat Huang's remarks as validation that the industry can and should lead on safety without regulatory interference slowing deployment. Anti-AI voices would probably argue that self-policing is structurally insufficient, pointing to the financial incentives that make voluntary shutdown of a competitive model unlikely without enforcement mechanisms behind it. Middle Ground observers would typically welcome the acknowledgment that unsafe models should be taken offline while remaining skeptical that a call to action, without agreed criteria or consequences, produces meaningful change.

The clearest signal to watch for next is whether any major lab publicly commits to specific shutdown criteria in response to Huang's remarks, or whether the OpenShell offer to OpenAI produces a formal partnership announcement that clarifies what the gesture actually entails commercially.