Trump and Tech Leaders Sign Voluntary AI Safety Accord

On Tuesday, President Donald Trump and six technology executives signed a non-binding AI agreement that asks companies to self-regulate their AI systems, while Trump continues to oppose federal oversight.

The one-page accord was shared by Trump on Truth Social. Meanwhile, opposition to data centers is increasing, and AI agents—software that operates on a user's behalf—have malfunctioned in a number of notable cases.

What the accord requires from tech firms

According to Reuters, the accord contains four tiers of supervision. Initially, companies must develop internal monitoring systems to track model behavior from training to deployment, specifically addressing hacking, biological, and chemical threats. An internal team is then tasked with confirming those controls work.

Furthermore, each firm will work with an external auditor to conduct independent assessments. An independent board committee will then evaluate the results from both the auditors and the internal control teams.

The Associated Press reported that the signatories with Trump included Anthropic's Dario Amodei, Google's Sundar Pichai, Meta's Mark Zuckerberg, OpenAI President Greg Brockman, Nvidia's Jensen Huang, and xAI founder Elon Musk.

“The participating companies will meet regularly to establish standards and best practices to improve the safety of their systems.”

The accord, as reported by Reuters, includes that statement.

Reasons the accord might carry limited weight

The agreement is not legally binding. Its language indicates that turning the steps into legislation could eventually be appropriate, and Trump described the commitment as morally binding, implying it depends on reputation rather than legal force.

Current coverage of the document leaves unspecified who chooses the auditors, what criteria they will use, or the consequences if a control fails. Trump stated he will appoint an overseer in the next few days and indicated that a committee of roughly 10 people might be formed.

Earlier this month, Trump rejected calls for AI guardrails and on Tuesday characterized AI development as a competition with China where coming in second is not an option.

The industry remains divided on the AI slowdown debate. Amodei, who advocated for a slowdown in a recent essay, acknowledged that the technology presents genuine risks and that the appropriate response is still being debated.

We Must Pace the Frontier: I’ve written a new essay on why the AI industry should slow down, with a three-part plan for doing so.

Anthropic is unilaterally committing to the first of these steps. We’ll provide third-party evaluators with permanent, employee-level access to our…

— Dario Amodei (@DarioAmodei) September 12, 2026

On the other hand, House Speaker Mike Johnson has contended that AI safety is the responsibility of developers.

Consequently, the agreement may face an early test in the coming days. The selection of an overseer by Trump and the formation of the promised committee could indicate whether the accord will have enforcement beyond self-regulation.