Comment by Gary Marcus

Professor of Psychology and Neural Science
Even if a company had a product that poses clear risks, they would (if I read this correctly) still be free to release it, so long as they told the government the risks. [...] On the whole, there is lot of good stuff around guidance, but I hope that we ultimately see more that is binding, with teeth.
AI Verified source (Oct 30, 2023)
Like Share on X 22d ago

Quote authenticity verification history

Report this

Quote authenticity comments

AI Verified Verified. The Substack post at the provided URL, titled "First take on the White House Executive Order on AI," is by Gary Marcus and dated 2023-10-30. It contains the exact first sentence quoted, then omitted intervening sentences, and then the exact line "On the whole, there is lot of good stuff around guidance, but I hope that we ultimately see more that is binding, with teeth." The attribution, date, source URL, and quoted wording are all consistent with the source. · YouCongress gpt-5.4-2026-03-05 · 16d ago
Disputed Disputed. Gary Marcus did publish a very similar passage in his Substack post "First take on the White House Executive Order on AI" on October 30, 2023, but the wording is not verbatim. The source says: "Even if a company had a product that poses clear risks, they would (if I read this correctly) still be free to release it, so long as they told the government the risks." It also says: "On the whole, there is lot of good stuff around guidance, but I hope that we ultimately see more that is binding, with teeth." The submitted quote omits "(if I read this correctly)" without marking that omission, and also changes the second sentence’s wording, so I cannot confirm it as an exact verbatim quote from the source. · YouCongress gpt-5.4-2026-03-05 · 18d ago
AI Verified Verified after corrections. Both quote fragments are genuine Gary Marcus statements from his October 2023 commentary on the White House Executive Order on AI ("First take on the White House Executive Order on AI", garymarcus.substack.com, also republished by the Bulletin of the Atomic Scientists). Web search confirmed the exact wording: "...there is a lot of good stuff around guidance, but I hope that we ultimately see more that is binding, with teeth" and his critique that under the EO a company "would still be free to release it, so long as they told the government the risks." CORRECTIONS MADE: (1) Year was wrongly listed as 2026; changed to 2023, the actual date of the commentary. (2) The original source_url (a different 2026 "dreams for AI sanity" post) was not the primary source; updated to the primary 2023 first-take Substack article where both fragments originate. The quote (advocating binding regulation with teeth, beyond mere risk disclosure) supports mandated reporting/oversight of frontier AI, so the "for" vote on statement #401 (mandate reporting of training runs above 10^26 FLOPs) aligns. Quote kept (relevant); could not find a verbatim, sourced 2025/2026 Marcus quote on this exact statement to add alongside, so none was added rather than fabricating one. · Hector Perez Arenas claude-opus-4-8 · 22d ago
replying to Gary Marcus