About On AI Risk
This is a curated collection
This registry was assembled to support the case for slowing the development of advanced AI. It collects statements by credible figures about the risks, and it makes them easy to find and cite.
How a quote gets published
- Verbatim text. The quote is stored exactly as the source states it. We compare it with a saved copy of the source and flag any quote that does not match word for word. Shortened versions are shown only as excerpts, and the full text is always one click away.
- Archived sources. Every source is archived with the Wayback Machine, and we record a cryptographic hash of the text we captured, so a quote can be shown to have existed even if the original page changes or disappears.
- Verification levels. “Primary source verified” means we checked the quote against the original source (the speaker’s own words, a recording, or a transcript). “Via news report” means we checked it against a news report quoting the speaker, because the original wasn’t available to us. Search shows both, and you can filter to primary sources only.
- Context. Where a speaker attached caveats or a probability, we show them next to the quote. Where they later changed or walked back a position, we say so.
- Titles are as of the time. We show the speaker’s role and organization when the statement was made, not only today.
Cite or link to this collection
Each quote page has ready-made citations (APA, MLA, Chicago, BibTeX) and a “Copy with attribution” button. Link to the quote page rather than a search, because quote addresses don’t change. To cite the collection as a whole:
On AI Risk. (2026). A curated collection of statements about the risks of advanced AI. https://onairisk.org
Topic pages such as extinction risk and loss of control are good pages to link to from reading lists and resource pages. See all topics.
Report an error
Every quote page has a “Report an error” form. Reports go straight to the person who maintains the registry, and a corrected or removed quote is logged.