Story
October 1, 2026
Trump’s AI safety deal leaves Big Tech to police itself
The White House and AI executives portray their voluntary accord as a workable route to safer frontier technology without choking off innovation. Critics see the same deal as an endorsement of self-regulation at a moment when increasingly capable systems are already testing the limits of corporate oversight.
Recent security incidents and pauses in frontier-model development had intensified the argument over whether AI companies could reliably contain the systems they are racing to build. The pressure created an opening for the White House to seek a common safety framework—without imposing binding federal rules.1
On Tuesday, President Donald Trump convened leaders from major AI firms at the White House and announced the voluntary White House Accord on Super Intelligence. The document asks participating companies to monitor models for cyber, biological and chemical risks; empower internal safety teams; use external auditors or evaluators; and create board committees to oversee the process. It also says the companies will meet to develop common standards and best practices.2
Trump called the arrangement “morally binding” and described it as “a constitution, in a way,” framing self-policing as protection while insisting the United States must not stifle growth. The pact’s signatories included leaders from Google, Anthropic, Meta, OpenAI, xAI and Nvidia.3
The companies’ public response was upbeat. Meta’s Mark Zuckerberg called the commitment to “robust internal controls and multiple layers of audits and reviews” a “significant positive step,” saying it should build confidence that the technology works as intended.
4 Yet even supporters described it as a beginning, not a final answer: Zuckerberg said the accord was “a start” rather than the only action industry would ever take.5
Critics argue that is precisely the weakness. The agreement sets no legal penalties, leaves key terms such as “robust” and “operating as intended” undefined, and permits companies to select the outside evaluators meant to scrutinize them. In that view, neither board committees nor paid auditors substitute for public transparency and enforceable rules.6
The divide is now stark: the administration sees coordinated industry responsibility; skeptics see the firms most exposed to AI risk writing—and grading—their own homework.