Claim · 2022-12-15 · publication · Stated in source · Supported

The paper describes a method it calls Constitutional AI (CAI) and its use to train a non-evasive and relatively harmless AI assistant without human feedback labels for harms.

One recorded claim. The claim panel gives its qualification, the source passages with their place in each source, and how it was checked.

What it connects

Constitutional AI: Harmlessness from AI Feedback → states its contribution → The paper describes a method it calls Constitutional AI (CAI) and its use to train a non-evasive and relatively harmless AI assistant without human feedback labels for harms.

Where it appears

No collection page holds this claim. It is listed with its records in the explorer.

Claim and evidence

Constitutional AI: Harmlessness from AI Feedbackstates its contribution ↓The paper describes a method it calls Constitutional AI (CAI) and its use to train a non-evasive and relatively harmless AI assistant without human feedback labels for harms.

Machine-checked, not reviewed by an editorStated in sourceSupported

Supporting evidence: primary · 1 source

Date: 2022-12-15 · publication Records dated 2022 →

Scope and limitsThe paper's own description of its contribution; not an assessment of it.

The evidence 01

primary sourceSupports the claim

Constitutional AI: Harmlessness from AI Feedback

arXiv · 2022-12-15

Cite the original source

Cite the original document, not Selekoa. Check these details against it.

Source excerpt

“In this paper we develop a method we refer to as Constitutional AI (CAI), depicted in Figure 1, and use it to train a non-evasive and relatively harmless AI assistant, without any human feedback labels for harms.”

Location in source arXiv v1, PDF p. 2, §1 Introduction, para. 2

Machine-checked · 2026-10-31

Source details and check notes

Authors: Yuntao Bai; Saurav Kadavath; Sandipan Kundu; Amanda Askell; Jackson Kernion; Andy Jones; Anna Chen; Anna Goldie; Azalia Mirhoseini; Cameron McKinnon; Carol Chen; Catherine Olsson; Christopher Olah; Danny Hernandez; Dawn Drain; Deep Ganguli; Dustin Li; Eli Tran-Johnson; Ethan Perez; Jamie Kerr; Jared Mueller; Jeffrey Ladish; Joshua Landau; Kamal Ndousse; Kamile Lukosuite; Liane Lovitt; Michael Sellitto; Nelson Elhage; Nicholas Schiefer; Noemi Mercado; Nova DasSarma; Robert Lasenby; Robin Larson; Sam Ringer; Scott Johnston; Shauna Kravec; Sheer El Showk; Stanislav Fort; Tamera Lanham; Timothy Telleen-Lawton; Tom Conerly; Tom Henighan; Tristan Hume; Samuel R. Bowman; Zac Hatfield-Dodds; Ben Mann; Dario Amodei; Nicholas Joseph; Sam McCandlish; Tom Brown; Jared Kaplan
Version: arXiv:2212.08073v1, submitted 15 December 2022 (UTC). Retrieved for the candidate pack on 2026-10-28 (SHA-256 9a456a07ad346e3372f9867d346f69f5b0f68b4c65f060aca0b8a13fa9d98e83). The author list is the byline as printed on the first page.
Version date: 2022-12-15
Canonical address: Open the canonical source ↗
Accessed: 2026-10-28
Language: English (en)
Attribution: Yuntao Bai and others, “Constitutional AI: Harmlessness from AI Feedback”, 2022-12-15.
Reuse: No open redistribution licence asserted. Bibliographic metadata, editorial paraphrases and short attributed quotations (at most 50 words each). Recheck terms before wider redistribution.
Check notes: Candidate pack passage, key papers (research/key-papers). The quotation (37 words) matched the text retrieved for the candidate pack on 2026-10-28 (research/key-papers/verification.json). Matching ignored differences of whitespace, case, quotation marks, dashes and line-break hyphens, so the source’s typography can differ. No human editorial review.

Status and meaning

Editorial status. Not reviewed by an editor. An exploratory record that no editor has accepted. This status does not say whether a person has read any linked passage. Each evidence record above shows how it was checked.

Relation. A paper or other publication describes its own contribution, quoted from the expression the record names, usually its abstract or introduction. The record is the document’s own description, not an assessment: a claim of priority or novelty it contains is the document’s claim, flagged as such, and says nothing about earlier or later work.

Date. Date of the arXiv version quoted, submitted 15 December 2022 (UTC); later versions may word it differently.

Record ID kp-states-constitutional-ai-harmlessness-from-ai-2022-1