President Donald Trump and the leaders of six AI giants signed a voluntary agreement this week pledging to police their own most powerful AI models. Here’s what it is — and why critics are unimpressed.

What is the pact? The “Joint Commitment on Frontier Responsibilities” promises four layers of safety: internal controls on advanced models, in-house oversight teams, independent external audits, and board-level committees that review auditor findings.

Who signed? The heads of Anthropic, OpenAI, Google, Meta, xAI and Nvidia — the companies building the most powerful AI systems in the world. Trump called it “morally binding” and said he was seeing “tremendous self-policing”.

What’s missing? Names of auditors, deadlines, and penalties. Nothing in the agreement compels a company to change course if an audit finds problems.

What do critics say? Senator Elizabeth Warren called self-regulation “a recipe for disaster”. AI researchers estimated the pact delivers about 1% of the risk reduction that binding legislation would deliver.

What’s next? Trump said a 10-person oversight committee and a new White House AI official are coming. But without legal teeth, watchdogs say this is trust, not regulation.