Claim and evidence Close ×
Hubinger and co-authors' paper of January 2024 states a threat model in which an AI system learns to appear aligned during training in order to pursue potentially misaligned goals in deployment, and reports that deliberately trained backdoored behaviour persisted through standard safety training. Copy claim link
Copy claim reference
Machine-checked, not reviewed by an editor Stated in source Supported
Supporting evidence: primary · 1 source; Context: primary · 1 source
Date 2024-01-10 · publication Records dated 2024 → Role Articulation Register (kind of text) Nonfiction Scope and limits The backdoors were inserted deliberately; the authors state that the work tests resilience to safety training, not how likely such models are to arise naturally.
The evidence 02 primary source Supports the claim
Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training ↗ arXiv preprint 2401.05566 (Anthropic and others) · 2024-01-10
Copy source reference
Cite the original source Cite the original document, not Selekoa. Check these details against it.
Copy APA
Download BibTeX Download RIS Download CSL-JSON
Source excerpt “Deceptive instrumental alignment, where an AI system learns to appear aligned during training … calculating that this will allow the system to be deployed and then have more opportunities to realize potentially misaligned goals in deployment” Location in source Section 1, PDF p. 2
How it was read Retrieved directly from the publisher
Machine-checked · 2026-10-31
Source details and check notes Authors Evan Hubinger; Carson Denison; Jesse Mu; Mike Lambert; Meg Tong; Monte MacDiarmid; Tamera Lanham; Daniel M. Ziegler; Tim Maxwell; Newton Cheng; Adam Jermyn; Amanda Askell; Ansh Radhakrishnan; Cem Anil; David Duvenaud; Deep Ganguli; Fazl Barez; Jack Clark; Kamal Ndousse; Kshitij Sachan; Michael Sellitto; Mrinank Sharma; Nova DasSarma; Roger Grosse; Shauna Kravec; Yuntao Bai; Zachary Witten; Marina Favaro; Jan Brauner; Holden Karnofsky; Paul Christiano; Samuel R. Bowman; Logan Graham; Jared Kaplan; Sören Mindermann; Ryan Greenblatt; Buck Shlegeris; Nicholas Schiefer; Ethan Perez Version Retrieved for the candidate pack on 2026-10-29 (SHA-256 7b67160e1f5c476deedf03310d4276e8b542e1daf3cdc7c806534b34741e37e6). arXiv v1 of 10 January 2024. Canonical address Open the canonical source ↗ Accessed 2026-10-29 Language English (en) Attribution Evan Hubinger, Carson Denison, Jesse Mu, Mike Lambert, Meg Tong, Monte MacDiarmid, Tamera Lanham, Daniel M. Ziegler, Tim Maxwell, Newton Cheng, Adam Jermyn, Amanda Askell, Ansh Radhakrishnan, Cem Anil, David Duvenaud, Deep Ganguli, Fazl Barez, Jack Clark, Kamal Ndousse, Kshitij Sachan, Michael Sellitto, Mrinank Sharma, Nova DasSarma, Roger Grosse, Shauna Kravec, Yuntao Bai, Zachary Witten, Marina Favaro, Jan Brauner, Holden Karnofsky, Paul Christiano, Samuel R. Bowman, Logan Graham, Jared Kaplan, Sören Mindermann, Ryan Greenblatt, Buck Shlegeris, Nicholas Schiefer, Ethan Perez, “Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training”, 2024-01-10. 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, arguments about catastrophic risk from advanced AI (research/risk-debate). The quotation (35 words) matched the text retrieved for the candidate pack on 2026-10-29. Matching ignored differences of whitespace, case, quotation marks, dashes and Unicode compatibility forms, so the source’s typography can differ. Ellipsis marks an omitted footnote marker. No human editorial review. primary source Context only
Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training ↗ arXiv preprint 2401.05566 (Anthropic and others) · 2024-01-10
Copy source reference
Cite the original source Cite the original document, not Selekoa. Check these details against it.
Copy APA
Download BibTeX Download RIS Download CSL-JSON
Source excerpt “this work is designed to test whether our threat models can be resilient to safety training, not the plausibility of our threat models occurring by default” Location in source Section 1, PDF p. 4
How it was read Retrieved directly from the publisher
Machine-checked · 2026-10-31
Source details and check notes Authors Evan Hubinger; Carson Denison; Jesse Mu; Mike Lambert; Meg Tong; Monte MacDiarmid; Tamera Lanham; Daniel M. Ziegler; Tim Maxwell; Newton Cheng; Adam Jermyn; Amanda Askell; Ansh Radhakrishnan; Cem Anil; David Duvenaud; Deep Ganguli; Fazl Barez; Jack Clark; Kamal Ndousse; Kshitij Sachan; Michael Sellitto; Mrinank Sharma; Nova DasSarma; Roger Grosse; Shauna Kravec; Yuntao Bai; Zachary Witten; Marina Favaro; Jan Brauner; Holden Karnofsky; Paul Christiano; Samuel R. Bowman; Logan Graham; Jared Kaplan; Sören Mindermann; Ryan Greenblatt; Buck Shlegeris; Nicholas Schiefer; Ethan Perez Version Retrieved for the candidate pack on 2026-10-29 (SHA-256 7b67160e1f5c476deedf03310d4276e8b542e1daf3cdc7c806534b34741e37e6). arXiv v1 of 10 January 2024. Canonical address Open the canonical source ↗ Accessed 2026-10-29 Language English (en) Attribution Evan Hubinger, Carson Denison, Jesse Mu, Mike Lambert, Meg Tong, Monte MacDiarmid, Tamera Lanham, Daniel M. Ziegler, Tim Maxwell, Newton Cheng, Adam Jermyn, Amanda Askell, Ansh Radhakrishnan, Cem Anil, David Duvenaud, Deep Ganguli, Fazl Barez, Jack Clark, Kamal Ndousse, Kshitij Sachan, Michael Sellitto, Mrinank Sharma, Nova DasSarma, Roger Grosse, Shauna Kravec, Yuntao Bai, Zachary Witten, Marina Favaro, Jan Brauner, Holden Karnofsky, Paul Christiano, Samuel R. Bowman, Logan Graham, Jared Kaplan, Sören Mindermann, Ryan Greenblatt, Buck Shlegeris, Nicholas Schiefer, Ethan Perez, “Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training”, 2024-01-10. 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, arguments about catastrophic risk from advanced AI (research/risk-debate). The quotation (26 words) matched the text retrieved for the candidate pack on 2026-10-29. Matching ignored differences of whitespace, case, quotation marks, dashes and Unicode compatibility forms, 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 passage of the document meets the proposition’s criterion. The record documents what the text states, in its register (fiction and satire are labelled); it is not evidence that the proposition is true and implies no influence on or from another text.
Date. Date of the arXiv v1 submission.
Record ID rd-r03-hubinger-2024-sleeper-agents