Please share this with anyone doing AI research with 3rd party providers so that they can ensure their research won’t be corrupted.
- When you ask OpenRouter[1] to give you tokens from a given model, OpenRouter sends your request to a random available provider.
- OpenRouter providers have variable quality. Ensuring that your provider is high quality is really difficult.
- There is precedent for an AI safety paper accepted to NeurIPS having its core results entirely overturned by these issues.
- A review of influential AI Safety research codebases that use OpenRouter for their reported results found that 31/32 (97%) of them use OpenRouter unsafely.[2]
- Researchers who wish to do research using OpenRouter or similar providers should take precautions to minimize the risks to their research,[3] though the current selection of providers is insufficient for ideal scientific reliability.
- Replicators should test whether results hold up when these bugs are fixed.
- Core AI Safety codebases should make fixes so that downstream users are able to implement best practices.
This is a tangent I took for a few days during the Pivotal AI Safety Research Fellowship. I’m doing an AI Control project mentored by Adam Kaufman and James Lucassen of Redwood Research and partnered with Aniruddh [...]
---
Outline:
(02:04) OpenRouter Kills an AI Safety NeurIPS Paper
(03:29) Background
(04:15) Prior work/writing on the subject.
(05:38) No Guarantees from Fixing a Provider
(06:14) No Guarantees From Large N
(07:09) No Guarantees from "The Benchmark Didn't Move"
(08:19) No Guarantees from "Choose the Most Expensive Provider"
(08:54) No Guarantees from "Set a floor"
(09:12) When is Default OpenRouter Fine?
(09:50) Developing Best Practices
(11:45) Spotting Fingerprints
(13:12) Highlighting Some Impacted Work
(13:57) en-US-AvaMultilingualNeural__ Infographic titled "Provider Routing Inspector" with statistics and bar chart on OpenRouter reliability.
(14:10) Should The Community Do Something About This?
(14:15) Fixing Repos
(14:34) Replicating Research
(15:01) A New Org?
The original text contained 12 footnotes which were omitted from this narration.
---
First published:
July 23rd, 2026
---
Narrated by TYPE III AUDIO.
---
Images from the article:
Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.
Fler avsnitt av LessWrong (30+ Karma)
Visa alla avsnitt av LessWrong (30+ Karma)LessWrong (30+ Karma) med LessWrong finns tillgänglig på flera plattformar. Informationen på denna sida kommer från offentliga podd-flöden.
