Self-Policing, Internal Checks: Inside Trump’s Accord on ‘Super Intelligence’

US President Donald Trump and executives from some of the world's largest artificial intelligence companies have signed a voluntary agreement focused on the development and safety of what the administration calls “super intelligence” (SI).

The White House Accord on Super Intelligence: A Joint Commitment on Frontier SI Responsibilities was unveiled on September 29 after Trump met technology executives at the White House. The document sets out a framework under which companies developing advanced AI systems would take responsibility for monitoring and managing safety risks through their own internal processes, external assessments and board oversight.

What is the ‘Super Intelligence’ accord?

The one-page accord is a voluntary commitment, rather than a new federal law or regulatory regime. Trump described it as “morally binding”, while the document says its principles could potentially be incorporated into laws or regulations in the future.

The agreement was signed by Trump and six technology executives: Google CEO Sundar Pichai, Anthropic CEO Dario Amodei, Meta CEO Mark Zuckerberg, OpenAI President Greg Brockman, Nvidia CEO Jensen Huang and xAI founder Elon Musk.

Four layers of oversight

The accord establishes four main layers of controls for participating companies.

1. Robust internal controls:
Companies are expected to monitor the capabilities and alignment of their AI models during training and deployment. The controls are intended to address risks including cybersecurity, biosecurity and chemical threats.

2. Internal evaluation teams:
Companies would maintain internal teams responsible for assessing whether their safety controls are functioning and identifying problems that require attention.

3. Independent external audits:
The companies would work with independent external auditors or evaluators to assess whether their internal safety measures and monitoring systems are working as intended.

4. Independent board committees:
Each participating company would establish an independent committee at board level to review reports from internal teams and external auditors.

The companies also agreed to work together on safety standards and best practices.

Trump backs industry self-policing

Trump has emphasised that the agreement relies heavily on industry self-regulation rather than direct government controls.

“I think I'm seeing tremendous self-policing,” Trump said after meeting the executives. He has argued that excessive government restrictions could slow technological and economic development.

Nvidia CEO Jensen Huang similarly said there was no inherent conflict between AI innovation and safety, arguing that technological advances could also be used to develop better AI-safety mechanisms.

Why the agreement matters

The accord comes amid growing debate in the US over how governments should respond to increasingly capable AI systems.

AI companies have already developed internal safety and risk-management procedures, meaning several provisions in the agreement formalise practices that some firms were already using. The Washington Post reported that the commitments fall short of the level of government intervention sought by some technology experts and lawmakers.

Anthropic CEO Dario Amodei said the agreement was intended to give the public greater confidence that AI systems would operate as intended, while describing the deal as a starting point.

Trump changes terminology to ‘super intelligence’

The White House announcement was accompanied by Trump's decision to use “super intelligence” rather than “artificial intelligence” in federal terminology.

Trump said the new terminology better reflected the advanced capabilities of the technology. The move has itself become part of the wider debate over how the rapidly developing technology should be described and governed.

Government oversight remains a separate question

The voluntary nature of the accord means it does not itself establish enforceable federal safety requirements for the participating companies.

Trump has nevertheless discussed creating a roughly 10-member committee to oversee developments in the AI sector and has also indicated that he intends to appoint an AI czar to coordinate government initiatives.

The accord therefore represents a framework built around company-level controls, independent assessment and board oversight, while the broader question of how much statutory government regulation should apply to frontier AI remains under debate.