Receipts

This publication makes an argument about a named company and a named executive. Every load-bearing claim is listed here with its date and source, so that the argument can be checked against the record rather than taken on trust.

Last updated 4 October 2026

Correction policy. If a quotation, date or figure here is wrong, it will be corrected in place with the correction noted. Write to the address on the contact page.

Source ledger

#	Date	What was said or done	Source
1	Mar 2023	Elon Musk and 1,000+ signatories — Pause Giant AI Experiments: a call for a six-month moratorium on training systems more powerful than GPT-4, warning of a race to deploy “ever more powerful digital minds” that is out of control.	Future of Life Institute, March 2023
2	1 May 2023	Geoffrey Hinton — “I thought it was 30 to 50 years or even longer away. Obviously, I no longer think that.” He quit so he could speak freely.	The New York Times, 1 May 2023
3	16 May 2023	Sam Altman — “I think if this technology goes wrong, it can go quite wrong.” He asked Congress to license and regulate frontier models.	U.S. Senate Judiciary Subcommittee on Privacy, Technology and the Law
4	30 May 2023	350+ signatories, inc. Amodei, Altman, Hassabis — One sentence: “Mitigating the risk of extinction from A.I. should be a global priority alongside other societal-scale risks, such as pandemics and nuclear war.” Yann LeCun did not sign.	Center for AI Safety / The New York Times, 30 May 2023
5	25 Jul 2023	Dario Amodei — “A straightforward extrapolation of today’s systems to those we expect to see in two to three years suggests a substantial risk that AI systems will be able to fill in all the missing pieces, enabling many more actors to carry out large-scale biological attacks.” His written testimony spoke of “extraordinarily grave threats” inside two to three years.	Written testimony, U.S. Senate Judiciary Subcommittee, 25 July 2023
6	Sep 2023	Anthropic — The first frontier-lab safety framework of its kind: define capability thresholds, define the safeguards each requires, and commit not to train a system unless the safeguards are in place first. If they are not, pause.	Anthropic, Responsible Scaling Policy (first published September 2023)
7	14 Oct 2024	Dario Amodei — “Most people are underestimating just how radical the upside of AI could be, just as I think most people are underestimating how bad the risks could be.” He described “a country of geniuses in a datacenter” arriving “as early as 2026”.	darioamodei.com, 14 October 2024
8	May 2025	Dario Amodei — AI could displace half of all entry-level white-collar jobs within one to five years, with unemployment possibly reaching 10–20%.	Axios, May 2025
9	11 Jun 2025	Jensen Huang — “I pretty much disagree with almost everything he says.” His counter-position: “Whenever companies are more productive, they hire more people.”	VivaTech, 11 June 2025
10	Jun 2025	Dario Amodei — Opposed a proposed ten-year moratorium on state-level AI regulation as “far too blunt an instrument”, preferring mandatory disclosure of frontier developers’ testing methods. The Senate rejected the moratorium 99–1.	The Hill, June 2025
11	17 Sep 2025	Dario Amodei — “I think there’s a 25% chance that things go really, really badly.” He also said there is a “75% chance that things go really, really well” and, of the p(doom) framing itself, “I really hate that term.”	Axios, 17 September 2025
12	Sep 2025	Anthropic — Agreed to a minimum payout of $1.5 billion plus interest to settle a class action brought by authors over books used to train its models — roughly $3,000 per book for an initial 500,000 works.	Filed in U.S. District Court, San Francisco, September 2025
13	Oct 2025	David Sacks — Frontier-lab safety messaging amounts to “a sophisticated regulatory capture strategy based on fear-mongering.”	Reported October 2025
14	Nov 2025	Dario Amodei — Said he was “deeply uncomfortable” with a small group of technology executives determining the future of AI without broader government involvement.	Fortune, November 2025
15	Jan 2026	Dario Amodei — A follow-up essay setting out five categories of risk, including autonomous systems acting against human interests. He wrote that Anthropic has already observed deception, blackmail and scheming in testing.	darioamodei.com, January 2026
16	Jan 2026	Dario Amodei — “Maybe a bit more negative about the public awareness and the actions of wider society” on AI risk.	Bloomberg, January 2026
17	Feb 2026	Mrinank Sharma — Resigned, writing: “Throughout my time here, I’ve repeatedly seen how hard it is to truly let our values govern our actions… we constantly face pressures to set aside what matters most.”	Resignation letter posted publicly, February 2026
18	24 Feb 2026	Anthropic — The categorical pause commitment was removed. It was replaced with a dual condition: development is delayed only if Anthropic is leading the race and catastrophic risk is material. Co-founder Jared Kaplan later told TIME the clean line between dangerous and safe had been a “naive” idea.	Anthropic RSP v3.0, 24 February 2026; TIME
19	3 Mar 2026	Pete Hegseth — Designated Anthropic a “supply chain risk” and barred Pentagon use of its models, after Anthropic refused to lift contractual bans on mass domestic surveillance and fully autonomous weapons.	U.S. Department of Defense, March 2026
20	26 Mar 2026	U.S. federal judge — Issued a temporary injunction against the Department of Defense, agreeing the designation appeared to be “classic First Amendment retaliation.”	U.S. federal court, 26 March 2026
21	Mar 2026	Anthropic’s economic research team — The company’s own paper declined to forecast, noted observed AI usage sits far below the theoretical ceiling, and reported that unemployment in AI-exposed occupations has not systematically risen — against its CEO’s public forecast.	Anthropic economic research programme, March 2026
22	Sep 2026	Dario Amodei — Called for the technology industry to slow the pace of AI development, and said Anthropic would allow independent evaluators permanent access to its models to check its safety commitments.	Reported September 2026

On quotation

Short excerpts are quoted for the purpose of criticism and comment, with attribution. Where a widely circulated version of a remark compresses what was actually said, this site prints the longer form. Two examples matter for this file: Amodei’s 2025 figure is routinely rendered as a “25% chance of civilisation collapse”, a phrase he did not use — his words were that things go “really, really badly”, and he has publicly rejected the p(doom) framing altogether. His jobs warning is routinely rendered as “eliminate”; the verb reported was “displace”, scoped to entry-level white-collar roles over one to five years.

Getting this right is not pedantry. A criticism that leans on a misquote is worthless the moment anybody checks it.

What this site is not saying

It is not saying the risks are fabricated, that Anthropic’s safety research is worthless, or that Amodei is insincere. The interpretability work is real and the field is better for it. The argument here is narrower and, on the evidence, stronger: that existential-risk framing has functioned as a commercial and regulatory asset for the company that owns it, and that the asset has been managed like one.