Google, Anthropic, Meta, OpenAI, xAI and Nvidia signed a voluntary AI safety agreement at the White House on Tuesday, September 29, committing to internal risk controls, independent evaluation and board oversight as President Donald Trump backed industry self-regulation.
Trump described the White House Accord on Super Intelligence as morally binding rather than legally enforceable. “They’re going to police themselves,” he told reporters after meeting technology executives.
The accord sets out four layers of oversight for companies training and deploying frontier models. Internal controls would monitor capabilities and alignment during training and deployment, including cybersecurity, biological and chemical risks. The document specifically calls for controls to prevent models from hacking or gaining unintended access to technical systems.
A separate internal team would check whether those controls and detection systems are working and whether problems are corrected. An independent external auditor or evaluator would assess the same systems. An independent board committee would receive reports from the operational teams and internal and external reviewers, and oversee remediation.
The participating companies also agreed to meet regularly to develop standards and best practices. The document leaves open the possibility that the measures could eventually be written into law or regulation.
The signature page names 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, alongside Trump. Other executives attended the meeting, but these six companies signed the accord.
Zuckerberg described the agreement as a starting point for industry-wide standards. Amodei said the technology carries real risks and that the mechanism for addressing them remains under discussion.
The meeting followed Trump’s rejection of calls to slow AI development earlier this month. Anthropic and OpenAI had separately promised outside safety evaluators employee-like access to their systems. Those earlier commitments described access arrangements; Tuesday’s accord sets out a common structure for company oversight.
A separate executive order signed Tuesday directs federal agencies, where legally permitted, to use “Super Intelligence” or “SI” instead of “artificial intelligence” or “AI” in public communications and other nonstatutory documents. It does not require agencies to rewrite previously issued regulations, contracts or historical records.
For the order’s purposes, the new term retains the existing statutory definition of artificial intelligence. The president’s science and technology adviser has 60 days to submit proposed legislative language for a federal definition of super intelligence, assess whether it should replace or expand the current definition, and recommend corresponding executive actions.
Sources: Trump’s Truth, AFP, White House
–
By the Control Plane Editorial Team