The Year of Governance
From the godfather's resignation to the Paris summit, governments and institutions tried to write rules for AI in real time. The frontier kept moving. By the time a rule was drafted, the technology it described had been superseded.
The alarm goes mainstream
Governments try to write rules
The pullback
Mar 22 · 2023
Future of Life Inst.
The "Pause" letter
A call for a 6-month halt on systems beyond GPT-4. Eventually ~33,000 signatures, a fake "Xi Jinping" entry, and no pause.
May 1 · 2023
New York Times
Hinton quits Google
The "Godfather of A.I." leaves so he can warn freely that his life's work may slip out of human control.
May 16 · 2023
U.S. Senate
Altman testifies
OpenAI's CEO asks Congress for federal licensing of frontier models. Critics call it a moat built of compliance.
May 30 · 2023
Center for AI Safety
22-word extinction statement
"Mitigating the risk of extinction from AI should be a global priority..." Signed by Hinton, Bengio, Altman, Hassabis, Amodei, Gates.
Jul 21 · 2023
White House
Voluntary commitments
7 companies (Amazon, Anthropic, Google, Inflection, Meta, Microsoft, OpenAI) agree to safety testing and watermarking. Non-binding.
Oct 30 · 2023
U.S. Executive Order
Biden EO 14110
100+ pages. Uses the Defense Production Act to require red-team reporting above 1026 FLOP.
Nov 1–2 · 2023
Bletchley Park, UK
First AI Safety Summit
28 countries + EU sign the Bletchley Declaration. China attends and signs. A series begins.
Dec 8 · 2023
European Union
EU AI Act agreed
First comprehensive AI law. Systemic-risk tier set by a compute threshold of 1025 FLOP.
Sep 29 · 2024
California
SB 1047 vetoed
Newsom rejects the frontier-model safety bill: regulating by model size, he argued, is regulating by the wrong variable.
Feb 10–11 · 2025
Paris
AI Action Summit
"Safety" dropped from the name. Vance attacks regulation; the U.S. and UK decline to sign the declaration.
23 months of rule-writing
From the EU's pre-ChatGPT draft (April 2021) to a rewritten, agreed law (December 2023) the technology it described had changed beyond recognition. The international consensus assembled at Bletchley in 2023 had visibly come apart by Paris in 2025. The gap between building and governing did not close. It widened.
SOURCE · NEURON MAKERS, CH. 28 & CH. 30 — drawn from research dossier 07 (NYT, CAIS, White House fact sheets, EU Council/Parliament releases, UK Bletchley Declaration, CA SB 1047 veto message, Paris AI Action Summit). Dates as published.