Trump Pairs Super Intelligence Order With Voluntary AI Self-Policing Pact
President Donald Trump on September 29, 2026, combined a new executive order on AI with a voluntary pledge from six leading AI executives to police their own frontier systems.
The order is titled 'Inaugurating the Era of Super Intelligence' The White House. A companion fact sheet carries the title 'Fact Sheet: President Donald J. Trump Inaugurates the Era of Super Intelligence'. Both were dated September 29 and listed in the White House briefing room under executive orders and fact sheets respectively.
The industry document is the Joint Commitment on Frontier Responsibilities. Its signatories are Google's Sundar Pichai, Anthropic's Dario Amodei, Meta's Mark Zuckerberg, OpenAI's Greg Brockman, xAI's Elon Musk and Nvidia's Jensen Huang The Verge. All six provided press commentary at Trump's behest praising the self-policing plan.
A voluntary accord
The commitment is described as a "morally binding" pledge requiring signatory companies to self-regulate their AI technology with common-sense guidelines. Trump said on September 29 that tech executives had agreed to establish voluntary standards for AI Reuters, and that the industry can police itself after signing the accord.
His formulation was direct. "I think I'm seeing tremendous self-policing. And they understand that they have to self-police" NPR. He made remarks about AI companies self-policing to reporters outside the West Wing PBS.
No federal rule followed. Trump dismissed concerns about lack of government AI guardrail enforcement by saying U.S. companies are "dominating China". He said "the smartest people in the world" are "watching over each other" instead of introducing federal AI regulations.
The meeting itself followed a familiar Washington sequence. Trump hosted a meal with Big Tech leaders on Tuesday before discussing his AI announcements. He held a meeting with tech executives on AI on September 29, 2026 Roll Call. On that day he also said top tech executives had signed a new commitment aimed at better self-policing their companies' AI CNN.
Renaming AI
Language was central to the rollout. Trump said he disliked the word "artificial" in AI because it sounds "fake" Business Insider. He separately said the word "artificial" isn't helpful NPR.
That rename has policy consequences. Trump previously described AI safety fears as a "scam" CNBC. He said his prior claim that AI safety fears were a "hoax" cooked up by Democrats no longer applies after renaming the technology to "Super Intelligence".
The administration placed the order in a longer sequence. In March 2026, the White House unveiled a National Super Intelligence Policy Framework, as recounted in the September fact sheet The White House.
What voluntarism means in practice
The broader context here is familiar to anyone who works with frontier models, evaluation harnesses, red-teaming pipelines, and staged deployment. Voluntary commitments can move quickly across pre-training, fine-tuning, and inference controls without rulemaking delays. They can also diverge quickly in implementation, because eval thresholds, disclosure norms, and incident reporting mean different things in different stacks.
In this author's view, the technical question to watch is not whether labs test, they do, but whether common-sense guidelines converge into interoperable practice. Worth flagging: model cards, system cards, third-party eval access, and post-deployment monitoring are where a pledge either becomes operational or stays rhetorical. Without shared definitions for capability thresholds and containment, self-policing risks producing six parallel safety regimes with limited comparability.
There is also a competitive logic at work. Frontier developers face simultaneous pressure to ship, to scale inference capacity, and to reassure enterprise buyers and regulators. A joint, morally binding statement lets them signal restraint while preserving architectural and product discretion. For practitioners, the useful test will be concrete. Which evals are published, which weights are withheld, which agentic capabilities require additional review, and how cross-lab information sharing actually functions when incidents occur.
My own experience watching my children grow up alongside successive platform shifts left me optimistic about adaptation, but attentive to gaps in oversight. Users adopt faster than institutions document. That pattern held for the PC, the web, mobile, and cloud. It is holding for generative systems. Voluntary regimes can narrow that gap if engineering teams treat safety work as release-blocking infrastructure rather than as communications.
What this enables, if implemented consistently, is continued scale-up of U.S. training and deployment under a light-touch federal posture. The administration is betting that peer scrutiny among lab leaders, plus market pressure from customers who require auditability and provenance, will substitute for codified guardrails while preserving speed against Chinese competitors.
