Gary Marcus boosts criticism of federal AI preemption bill as weakening catastrophe safeguards

GaryMarcus · x · 2026-10-18

Gary Marcus reposted Jeff Ladish and quoted a lengthy critique by safety researcher Nathan Calvin of a proposed US federal AI bill.

Key points:

The reviewer says the bill, while containing some sensible provisions, is "not remotely close to meeting the moment" and passing it would leave society less protected against catastrophic risks than under existing law.
The preemption provisions are far-reaching: they would override state laws like California's SB 53 and strip state courts of general powers to enjoin defendants and protect the public.
Ladish highlights the contradiction: Dario Amodei has urged governments to slow AI down, yet the bill would remove states' ability to do exactly that while making it very hard for the federal government to halt dangerous training runs.

More from Companies & People