--- title: "Evan Hubinger: everything in the record, dated and linked" description: "Evan Hubinger, Alignment researcher, Anthropic. Every dated statement and action, linked to its source, plus live headlines naming them." canonical: https://monitor.paciva.ai/person-evan-hubinger source: AI News Monitor publisher: Paciva license: CC BY 4.0 updated: 2026-11-11T12:27:15.376Z keywords: Evan Hubinger, Evan Hubinger AI, Evan Hubinger quotes, Evan Hubinger statements, Evan Hubinger on AI, what Evan Hubinger said about AI, AI leaders, AI news --- > Evan Hubinger, Alignment researcher, Anthropic. Every dated statement and action, linked to its source, plus live headlines naming them. Read on the web: https://monitor.paciva.ai/person-evan-hubinger The record behind this page is public and citable. Every figure has a dated row behind it, and the rows are at https://monitor.paciva.ai/api/ledger # Evan Hubinger Alignment researcher, Anthropic · Anthropic ## Why now No headline this week names Evan Hubinger. ## What they said 4 dated items: statements, actions and media appearances by Evan Hubinger, linked to their sources and ordered newest first. - Sep 9, 2026 In the record · Public statement, reported "Publicly concurred with Coxon's resignation statement, as reported by Axios; stated probability above 10 percent within the decade" [General AI](https://monitor.paciva.ai/topic-general) · [Axios](https://www.axios.com/2026/09/09/anthropic-insiders-warn-ai-could-kill-all-humans) - Dec 1, 2024 In the record "Discussed model organisms of misalignment research in a transcribed episode of the AXRP podcast." [Safety & Risk](https://monitor.paciva.ai/topic-safety) · [axrp.net](https://axrp.net/episode/2024/12/01/episode-39-evan-hubinger-model-organisms-misalignment.html) - Jan 10, 2024 In the record "Backdoor behavior can be made persistent, so that it is not removed by standard safety training techniques." [General AI](https://monitor.paciva.ai/topic-general) · [arxiv.org](https://arxiv.org/abs/2401.05566) - Jul 28, 2023 In the record "A capabilities evaluation is a model evaluation designed to test whether a model could do some task if it were trying to." [Industry Deals](https://monitor.paciva.ai/topic-deals) · [Model Releases & Benchmarks](https://monitor.paciva.ai/topic-models) · [greaterwrong.com](https://www.greaterwrong.com/posts/dBmfb76zx6wjPsBC7/when-can-we-trust-model-evaluations) [Correct this record](https://monitor.paciva.ai/notes?skind=correction&sctx=Evan%20Hubinger#suggest) ## Where they are named ## Stories naming Evan Hubinger ## Where the record shows more than one item on a topic Each line is one topic in date order. The Monitor shows the sequence and links every item; it draws no conclusion from it. - Jan 10, 2024 ["Backdoor behavior can be made persistent, so that it is not removed by standard safety training techniques."](https://arxiv.org/abs/2401.05566) - Sep 9, 2026 [Publicly concurred with Coxon's resignation statement, as reported by Axios; stated probability above 10 percent within the decade](https://www.axios.com/2026/09/09/anthropic-insiders-warn-ai-could-kill-all-humans) ## Nothing on file What a filing or a published statement says about money, employment or affiliation. Each row states the relationship, the date it is filed as of, and the document it comes from. Nothing here says why anyone did anything. No interest on file for Evan Hubinger. The Monitor records a disclosed interest only where a filing or the person's own published statement states one. An empty list means none has been recorded, not that none exists. Every political contribution and lobbying registration on this site is also listed together in [the ledger](https://monitor.paciva.ai/#ledger). ## Compare Evan Hubinger with someone else Pick a second person and read both records in date order, on one screen.