NeuralStackExecutives from Google, Anthropic, Meta, Nvidia, OpenAI and SpaceX’s xAI have signed the White House Accord on Super Intelligence with US President Donald Trump. The agreement sets out a four-layer control and audit system for artificial intelligence during training and deployment.
The accord calls on each company to establish internal controls that monitor the capabilities and alignment of its AI models. Those controls are also intended to help ensure that systems do not hack or access technical systems in unintended ways.
A second layer requires each signatory to create an internal team responsible for checking whether its controls, monitoring and detection systems operate as intended. That places ongoing oversight within each company rather than leaving safety review solely to outside organizations.
The companies also agreed to work with independent external auditors or evaluators. These assessments are designed to provide an additional review of the systems and controls established by the companies themselves.
The fourth layer calls for an independent committee of each company’s board to oversee the process. The committee would receive reports from internal teams as well as from outside auditors and evaluators.
Why this matters
The accord brings several of the most prominent AI companies under a shared framework for monitoring, evaluation and board oversight. Its structure links technical controls to internal accountability, independent assessment and corporate governance.
The companies further agreed to meet regularly to establish standards and best practices intended to improve the safety of their systems. Those discussions could shape how participating companies define effective monitoring and evaluation as their AI models are trained and deployed.
Implementation will now determine whether the four-layer approach becomes a meaningful common practice across the signatories. The accord’s emphasis on internal controls, outside evaluation and independent board committees gives the companies several mechanisms through which progress can be monitored.