What does the White House AI safety accord actually change?

Trump and the CEOs of Google, Meta, OpenAI, Anthropic, NVIDIA and xAI signed a voluntary 'Accord on Super Intelligence' promising safety controls and independent audits — a moral commitment, not law.

On September 29, President Trump and the CEOs of Google, Meta, OpenAI, Anthropic, NVIDIA and xAI signed a voluntary “Accord on Super Intelligence” at the White House, committing the big labs to internal safety controls, independent external assessments, and board-level oversight of advanced AI systems. The honest part first: it is not legally binding — Trump himself called it “morally binding” and said it may eventually be written into law. What it does change is expectations: every major lab has now publicly promised to run safety reviews, which gives researchers, journalists and future regulators a benchmark to hold them to.

What it does not change: no law, no product, no enforcement mechanism. The practical takeaway for everyday users: the big AI systems you use are being built by companies that just publicly pledged not to cut corners on safety — but since it is self-policing, treat the new wave of always-on AI agents (like the ones OpenAI and Meta are now shipping) like any powerful tool: know what data they can access before you turn them on.