Trump Administration Signs Voluntary AI Safety Agreement With Google, Meta, Others

On Sept. 29, 2026, the Trump administration reached a voluntary AI safety agreement with Google, Anthropic, Meta, OpenAI, xAI, and Nvidia. The accord was termed the White House Accord on Super Intelligence. The CEOs of those firms signed the accord: Sundar Pichai of Google, Dario Amodei of Anthropic, Mark Zuckerberg of Meta, Greg Brockman of OpenAI, Elon Musk of xAI, and Jensen Huang of Nvidia. Officials described the agreement as morally binding and said it does not carry the force of federal law or impose penalties for noncompliance.

Under the accord, companies pledged to implement internal monitoring of their AI models as the first layer of controls. They also committed to create an internal team to verify that those controls work properly, forming the second layer of the safety framework. The accord calls for the engagement of independent external auditors to review the models, constituting the third layer of oversight. A board-level committee will receive the auditors’ reports and oversee the process, serving as the fourth layer of governance.

The signatories also committed to meet regularly as part of the agreement to discuss AI safety. These meetings intend to set shared standards for AI safety among the participating companies. They will also focus on establishing best practices for AI safety through regular collaboration. Trump issued an executive order directing federal agencies to replace the term ‘AI’ with ‘Super Intelligence’ in official usage.

The order also requires those agencies to propose changes to the legal definition of the concept. The accord lacks both penalties and a timeline for implementation. The accord does not require companies to publish audit results. The accord does not define what makes an auditor independent.

Trump said the companies understood they needed to self-police their AI development efforts. He also said the document is morally binding for the executives who signed it. The agreement is voluntary and does not impose legal penalties for noncompliance. The accord’s reliance on voluntary means results in no government-imposed fines for non-adherence.

Trump’s executive order applies to all federal agencies, directing them to use the term Super Intelligence in official communications. The accord’s four-layer control system includes internal monitoring, internal verification, external audits, and board oversight. Companies agreed to meet regularly to develop shared AI safety standards and best practices. The voluntary nature of the agreement means that enforcement relies solely on the companies’ reputational incentives.

Trump’s self-police comment underscores that the executives are responsible for policing their own AI development. The CEOs signed the White House Accord on Super Intelligence on Sept. 29, 2026, as recorded in the announcement. The agreement does not impose any financial penalties for violations of its safety commitments. No set deadline exists for companies to implement the internal monitoring and auditing requirements.

Audit results are not required to be made public under the terms of the accord. The accord does not specify what makes an external auditor independent. The agreement is described as morally binding. It carries no force of federal law.

There are no penalties for noncompliance. No timeline exists for implementing the safety controls. Companies are not required to publish audit results. The accord does not define an independent auditor.

Companies pledged to implement internal monitoring of AI models. They also committed to create an internal verification team. The accord calls for external auditors to review the models. A board-level committee will oversee the auditors’ reports.

The signatories agreed to meet regularly to discuss AI safety. These meetings aim to set shared AI safety standards. They also aim to establish best practices for AI safety. Trump issued an executive order to replace the term AI with Super Intelligence.

The order requires agencies to propose changes to the legal definition. Trump said the companies understood they needed to self-police. He also said the document is morally binding for the signing executives.