Trump's AI agreement is strong on self-regulation, weak on enforcement
Orders federal agencies to rebrand AI as 'superintelligence'
Donald Trump has announced a voluntary agreement with Big Tech leaders to introduce safeguards for AI, describing the deal as "morally binding."
The agreement, called the Joint Commitment on Frontier Responsibilities, was unveiled following a White House luncheon with executives including Elon Musk, Mark Zuckerberg, Nvidia's Jensen Huang and Anthropic's Dario Amodei.
Trump said the agreement was "almost like a constitution," and would provide protection against the potential risks posed by advanced AI systems.
However, the two-page document appears to carry no legal force or enforcement mechanisms. Instead, it relies on companies to monitor their own technologies and oversee compliance with the proposed safety measures.
The announcement comes amid growing public concern about AI and calls from lawmakers, researchers and technology companies for stronger government oversight.
The technology industry has largely aligned itself with the Trump administration, lobbying against strict regulation on the grounds that it could undermine innovation and economic growth.
Four layers of oversight, no legal enforcement
The accord outlines four layers of controls and audits to help AI companies identify and manage risks as they develop new AI systems:
- The first requires companies to establish internal safety monitoring during the training of AI models.
- The second calls for dedicated internal teams to ensure those safeguards are working effectively.
- The third layer involves independent external auditors assessing the companies' safety controls.
- The fourth requires an independent board committee to review reports on those measures.
However, the agreement does not specify any role for government regulators, or require companies to publish the independent assessments’ findings.
It also appears to leave firms free to select their own auditors, appoint their own oversight boards and decide whether to disclose the results of their evaluations.
The commitments therefore rely heavily on voluntary compliance, rather than legally enforceable standards.
Industry golfclaps ‘better than nothing’ announcement
Michael Adjei, director of system engineering at Illumio, described the announcement as "better than nothing," but argued that self-regulation is not enough.
"Self-regulation is not the answer, especially as leaders of the major AI labs have themselves called for external regulation to enforce a universal slowdown so that safety can catch up," he said.
Amanda Brock, chief executive of OpenUK, said the announcement meant the US had avoided regulating AI, restricting innovation or banning open models.
"After months of concern, the good news is that the US has not regulated AI, restricted innovation or banned open models. It hasn't even imposed guardrails," she said.
Brock, who has historically spoken against regulation, said the agreement's emphasis on internal risk reviews, external auditors and independent board oversight amounted to self-policing, allowing open-weight AI models to continue expanding their adoption.
Federal departments to adopt 'superintelligence' terminology
Alongside the agreement, Trump signed an executive order directing federal departments and agencies to replace the terms "artificial intelligence" and "AI" with "Super Intelligence" and "SI" in official communications.
The order applies to government correspondence, public communications, websites, reports and policy documents, although it does not require changes to previously issued regulations or historical records.
It also calls for proposed legislative language to establish a federal definition of the new terminology.
The executive order is separate from the agreement signed by the technology executives.
A Quinnipiac University poll conducted between 24th and 27th September found that just 25% of respondents approved of Trump's handling of AI, while 58% disapproved.
The president’s opinion on AI is starkly opposed to that of his voters, for once. Despite his vocal support for the technology, a Pew Research survey earlier this year found that just 16% of Americans believe AI will have a positive impact on society.
Trump’s approval rating on AI was lower than his ratings on the Iran war, the economy and US-China policy in the Quinnipiac poll.