Daniela Amodei said Anthropic hoped to publish soon a paper that discusses the role government could play in funding academic AI safety research.

No deadline is given beyond 'soon'; judge against whether Anthropic published the described paper. True if Anthropic (as author or co-author) published a paper or policy report that discusses the role government could play in funding academic AI safety research. Mostly true if such a paper was published and discusses government funding of academic AI research with safety as a stated part, though not as its central topic. Mostly false if Anthropic published only related policy material (e.g. general government AI policy) that does not address government funding of academic safety research, or only discussed the idea in comments or testimony rather than a paper. False if positive evidence shows no such paper was published, e.g. Anthropic's publication list over the following years containing no such paper, or a report that the project was dropped.

Set before the research began.

1. E1 [PDF] RFI Responses: National AI Research Resource Interim Report - NSF NSF (hosting an Anthropic RFI comment submission) · 29 June 2022 Own record

   “The NAIRR represents an enormous opportunity for the U.S. Government to continue its longstanding support of academic research and advance AI development in the United States.”

2. E2 Charting a path to AI accountability Anthropic · 13 June 2023 Own record

   “Increase funding for interpretability research. Provide government grants and incentives for interpretability work at universities, nonprofits, and companies.”

3. E3 Third-party testing in AI policy Anthropic · 25 March 2024 Own record

   “It’s therefore crucial that governments create experimental infrastructure to help academic researchers test out and evaluate frontier AI systems, and develop their own frontier systems for beneficial purposes.”

The evidence shows Anthropic blog posts recommending government grants for interpretability work at universities [E2] and government infrastructure for academic researchers [E3], and an RFI comment supporting government backing of academic research without mentioning safety [E1]. None of the items shows that Anthropic published a paper discussing government funding of academic safety research, which the mostly true verdict requires.