Trump and Tech Companies Sign Voluntary Agreement on Oversight of Advanced AI
Donald Trump and representatives of OpenAI, Google, Meta, Anthropic, xAI and Nvidia signed a voluntary agreement calling for a four-tier oversight system for advanced AI models.
The Superintelligence Accord, signed on September 29 by President Donald Trump and representatives of leading U.S. AI companies, introduces a voluntary oversight framework for advanced models. The document is titled White House Accord on Super Intelligence and was signed by representatives of OpenAI, Google, Meta, Anthropic, xAI and Nvidia.
OpenAI President Greg Brockman signed the document for OpenAI, Sundar Pichai for Google and Mark Zuckerberg for Meta. The agreement calls for multilayered reviews of companies’ safety procedures, including an independent external evaluation.
The Superintelligence Accord Calls for Four Levels of Oversight
The agreement calls for four layers of oversight. The first is internal controls within companies, the second an internal team tasked with reviewing those controls. The third layer is an independent external auditor or evaluator, and the fourth an independent board committee.
The external evaluator is supposed to examine whether controls, monitoring and mechanisms for detecting problems operate as intended. The text therefore addresses not only technical model testing, but also the evaluation of processes through which companies monitor risks and respond to identified shortcomings.
The common framework covers companies that are among the significant developers and suppliers of AI systems in the United States. The agreement connects companies’ internal oversight, external evaluation and the role of their boards in a single stated model.
It Is Neither a Law nor a Binding Rule
White House Accord on Super Intelligence is a voluntary agreement, not a law or a regulatory rule with legal enforceability. The signatories state that the described procedures could be codified in laws or regulations in the future. In its current form, however, the document does not replace binding rules or government oversight.
The text also does not establish penalties for failing to meet the commitments. It does not set a uniform standard for how companies should select external auditors, nor does it require audit results to be published. It has also not been independently confirmed what specific access external evaluators will receive to the models, data and internal systems of individual companies.
Based on the available information, it is therefore not yet possible to determine whether the agreement will bring a measurable change to safety procedures beyond companies’ existing voluntary commitments.
What Comes Next
An important next step will be the publication of the agreement’s complete original text on the White House’s official website or directly by the signatories. To assess how the framework works in practice, the names of the selected auditors, the precise scope of their mandates and the dates of the first evaluations will be particularly important.
It also remains unclear whether companies will publish audit findings or measures adopted after problems are identified. Another significant development would be legislative or regulatory steps that turn the agreement’s voluntary procedures into obligations.
Sources
- Associated Press – Confirms the signing of the voluntary agreement, its signatories and the commitment to internal and external oversight.
- Washington Examiner – Publishes the agreement’s text, including independent external evaluation and board committee oversight.
- The White House – Confirms that the administration adopted a related official measure on September 29, 2026, concerning the designation of AI as “Super Intelligence.”
Verified and updated: 09/30/2026 15:21
