Tech leaders sign voluntary White House AI safety accord
Companies commit to outside audits and board oversight

President Trump and leaders from some of the country's largest technology companies have signed a voluntary agreement establishing safety controls for advanced artificial intelligence systems. Trump described the September 29 agreement as "morally binding" after a White House meeting with executives from companies including Meta, Google, Nvidia, Anthropic, OpenAI and Microsoft.
The accord establishes four layers of controls and audits for companies developing advanced AI. Companies commit to maintaining internal safeguards designed to ensure their systems operate as intended, assigning separate internal teams to evaluate those controls, using an independent external auditor or evaluator to assess whether safeguards are working, and establishing board-level oversight to review reports from both internal and outside evaluations.
The agreement also says companies should work to prevent their AI systems from hacking or accessing technical systems in unintended ways. Participating companies agreed to meet regularly to establish standards and best practices for improving AI safety. The document acknowledges that the voluntary framework could eventually become law, stating that some measures could eventually be codified through legislation or regulation.
The accord does not currently impose federal penalties for violations and relies largely on participating companies to implement and oversee the safeguards. Trump said the companies have strong incentives to police themselves because failures could put their businesses at risk. He also said the group discussed creating a 10-person committee to "watch over the enterprise" and that he expects to name an AI czar within several days. Details about the committee's membership, authority and the czar's role have not yet been announced.
Separately on September 29, Trump signed an executive order directing federal agencies to use "Super Intelligence" and "SI" instead of "artificial intelligence" and "AI" in official non-statutory government materials. The terminology change does not alter the underlying technology and differs from the established technical use of "superintelligence," which generally refers to hypothetical systems exceeding human capabilities across a broad range of tasks.