The White House and six major technology companies have signed a voluntary agreement introducing internal controls, independent audits and board-level oversight for companies developing and deploying frontier AI models. The agreement is intended to establish a common governance framework for advanced AI systems but does not create new legal obligations.
The document, formally called the White House Accord on Super Intelligence: Joint Commitment on Frontier Responsibilities, was signed by US President Donald Trump alongside executives from Google, Anthropic, Meta, OpenAI, xAI and NVIDIA. Signatories from the companies included Sundar Pichai, Dario Amodei, Mark Zuckerberg, Greg Brockman, Elon Musk and Jensen Huang.
At the centre of the accord is a four-layer system of controls for companies training and deploying frontier models. Companies are expected to establish internal processes that monitor model capabilities and alignment during both training and deployment. The document specifically mentions cybersecurity, biosecurity and chemical threats, as well as the risk of models gaining unintended access to technical systems.
A second internal team should verify that these safeguards, monitoring systems and detection mechanisms are operating as intended and ensure that identified problems are addressed. Companies are then expected to bring in an independent external auditor or evaluator to assess those controls.
The fourth layer moves oversight to board level. An independent committee of the board of directors should receive reports from the teams responsible for the controls, as well as from internal and external auditors and evaluators. That committee would also oversee the remediation of identified problems.
The participating companies have also committed to meeting regularly to develop standards and best practices for the safety of advanced AI systems. The accord states that these measures could eventually be incorporated into laws or regulations, although the current agreement itself is voluntary.
That leaves much of the implementation with the companies themselves. The accord does not establish a government enforcement mechanism, specify common technical thresholds for frontier models or prescribe a single audit methodology. It also does not define detailed requirements for publishing audit findings.
From AI to ‘Super Intelligence’
The agreement was announced on the same day that Trump signed an executive order directing US federal agencies to use the terms “Super Intelligence” and “SI” instead of “Artificial Intelligence” and “AI” in new official communications, reports and policy documents.
For now, however, the terminology change does not amount to a new technical definition of superintelligence. The executive order says that “Super Intelligence” initially refers to the technologies already covered by the existing US statutory definition of artificial intelligence. The White House science and technology adviser has been instructed to propose legislation within 60 days that could establish a separate federal definition.
The accord therefore applies to what the industry generally describes as frontier AI models rather than only to systems that have demonstrably surpassed human intelligence.
The agreement reflects the administration’s preference for industry-led oversight at this stage. Trump described the commitment as morally binding, while the framework itself leaves open the possibility that elements of the approach may later become part of formal regulation.
For AI developers, the immediate consequence is an explicit commitment by six of the largest companies in the sector to introduce independent evaluation and direct board responsibility into the governance of frontier models. How significant that becomes in practice will depend on the standards, audit procedures and disclosure requirements that the companies develop under the accord.
