Trump and AI giants sign self‑regulation accord instead of new US AI law

On September 29, President Donald Trump hosted top technology executives at the White House and announced a voluntary “accord” on frontier AI models instead of backing new federal legislation. Leaders from Meta, Nvidia, Anthropic, Amazon and other firms signed the agreement, which calls on each company to build robust internal safety teams, establish board‑level oversight, and submit their most powerful systems to independent external audits before broad deployment.

Crucially, the accord has no legal force: companies are effectively pledging to self‑police the risks posed by increasingly capable AI, including potential misuse for cyberwarfare or large‑scale disinformation. On Capitol Hill, several bills to create a dedicated AI regulator or mandate incident reporting remain stalled, as key Republicans warn against “over‑regulation” that could slow innovation. In the meantime, a patchwork of state‑level AI rules is emerging, increasing compliance complexity for large providers.

Supporters argue that the new framework at least sets common expectations around governance for “superintelligence‑class” models and can be updated faster than formal law. Critics counter that, without enforceable standards or penalties, it amounts to a political gesture that shifts responsibility back to companies. Whether this self‑regulatory experiment can prevent serious AI‑related harms will likely shape the next phase of US AI policy.