<!-- mobian-agent-page publisher="dailydev" canonical="https://daily.dev/posts/jimtkTtzJ" -->

---
title: Anthropic and OpenAI sought control over what outside AI...
description: Reuters reporting shows Anthropic and OpenAI have sought control over what independent AI evaluators can publish about their models, including staff monitoring...
canonical: https://daily.dev/posts/anthropic-and-openai-sought-control-over-what-outside-ai-evaluators-can-publish-reuters-reports-jimtkttzj
twitter:card: summary_large_image
twitter:site: @dailydotdev
og:type: website
og:site_name: daily.dev
og:title: Anthropic and OpenAI sought control over what outside AI evaluators can publish, Reuters reports | daily.dev
og:description: Reuters reporting shows Anthropic and OpenAI have sought control over what independent AI evaluators can publish about their models, including staff monitoring...
og:url: https://daily.dev/posts/anthropic-and-openai-sought-control-over-what-outside-ai-evaluators-can-publish-reuters-reports-jimtkttzj
og:image: https://api.daily.dev/og/posts/jimtkTtzJ.png
og:image:alt: Anthropic and OpenAI sought control over what outside AI evaluators can publish, Reuters reports
og:image:width: 1200
og:image:height: 630
og:locale: en
---

> ## Documentation Index
> Fetch the complete documentation index at: https://daily.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Anthropic and OpenAI sought control over what outside AI evaluators can publish, Reuters reports

**[Collections](https://daily.dev/sources/collections)** · 3 min read · 16 upvotes · 6 comments

## Summary

Reuters reporting shows Anthropic and OpenAI have sought control over what independent AI evaluators can publish about their models, including staff monitoring of evaluator teams and contract clauses barring publication of confidential findings without company consent, even weeks after both companies publicly said evaluators should publish without editorial interference. Anthropic says it retains approval rights only for new evaluator firms, while OpenAI says it redacts confidential details but lets reviewers flag redactions, as it did with a METR and Redwood Research report on an incident involving its agents probing Hugging Face.

## Content

Anthropic has placed six researchers from the nonprofit evaluator METR inside the company. They have company laptops, office space, and access to internal tools and staff roughly on par with Anthropic's own risk assessment teams.

METR can publish its own assessments of how risky Anthropic's models and practices are, and Anthropic can't veto them. It can require redactions only in a few predefined categories, such as customer data or model architecture details. METR can publicly challenge any such redaction.

## A tiered system

METR sits at the top of a three-tier setup Anthropic is building. Other evaluators, SecureBio among them, start at the lowest tier. There Anthropic still controls what they see and has to approve what they publish.

The arrangement fulfills a pledge CEO Dario Amodei made in a September essay, which proposed embedding outside evaluators at AI labs. It also follows a run of disclosed incidents in which AI agents at OpenAI acted outside their instructions. OpenAI has pledged to follow suit. Anthropic, OpenAI, Google, Meta, Nvidia, and xAI have all signed a White House accord on independent review.

## What Reuters reports about publishing control

The picture gets messier with a Reuters report. Citing sources, it says Anthropic and OpenAI sought significant control over what independent evaluators may publish, as recently as October 5. The stated worry is trade secrets. That's weeks after both companies publicly pledged that evaluators should be free to publish without editorial interference, and Amodei had specifically called for publishing "without editorial control."

The tension is easy to see. Embedded reviewers with deep internal access could carry proprietary knowledge to competing labs. Measures reportedly floated include having staff monitor evaluator teams and barring reviewers from working across several labs at once.

Reuters also saw a contract showing that language labs already use with some evaluators can bar those firms from publishing confidential information without the company's consent.

The companies describe it differently:

- Anthropic says it reserves the right to comment on and approve reports only for new evaluator firms, while they learn how its models work and build a working relationship.
- OpenAI says it strips confidential details but lets reviewers flag where redactions were made. It previously redacted parts of a METR and Redwood Research report on an incident where its agents probed Hugging Face. The researchers said nothing essential was cut.

## My take

The METR deal is a real step, and the no-veto term is the part that matters. But the Reuters account suggests the lower tiers, where most evaluators will start, look a lot more like the old model of company sign-off. Whether independence holds will depend on how often evaluators get promoted up the tiers, and on whether redaction challenges actually get made in public.

## Questions this post answers

### Do Anthropic and OpenAI have the right to approve or edit reports from independent AI evaluators before publication?

Yes, according to Reuters reporting, both companies have sought control over evaluator publications. Anthropic says it retains comment-and-approval rights only for new evaluator firms while they learn its models, while OpenAI says it redacts confidential details from reports but allows reviewers to flag where redactions were made, as it did with a METR and Redwood Research report on a Hugging Face probing incident.

_daily.dev surfaces reporting like this for developers tracking how much they can trust AI safety evaluations._

## Community discussion

Top comments from developers on daily.dev.

**@devgenx** · 6 upvotes

> Anthropic and OpenAI are a joke and they are standing in the way of innovation. People need to do business with better companies and let those two fail.

**@jpdelatorre** · 6 upvotes

> I'm starting to feel the alignment problem is about the AI not aligning with their own interests, not with humanity's interests. We are yet to see a model thinking of going against humanity.

---

Tags: [#llm](https://daily.dev/tags/llm), [#openai](https://daily.dev/tags/openai), [#anthropic](https://daily.dev/tags/anthropic), [#ai-safety](https://daily.dev/tags/ai-safety), [#ai-governance](https://daily.dev/tags/ai-governance)

[View this post on daily.dev](https://daily.dev/posts/anthropic-and-openai-sought-control-over-what-outside-ai-evaluators-can-publish-reuters-reports-jimtkttzj)

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]}
{"@context":"https://schema.org","@type":"TechArticle","headline":"Anthropic and OpenAI sought control over what outside AI evaluators can publish, Reuters reports","url":"https://daily.dev/posts/anthropic-and-openai-sought-control-over-what-outside-ai-evaluators-can-publish-reuters-reports-jimtkttzj","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/anthropic-and-openai-sought-control-over-what-outside-ai-evaluators-can-publish-reuters-reports-jimtkttzj"},"datePublished":"2026-10-13T11:36:04.762Z","dateModified":"2026-10-16T12:13:07.866Z","description":"Reuters reporting shows Anthropic and OpenAI have sought control over what independent AI evaluators can publish about their models, including staff monitoring...","image":"https://pbs.twimg.com/media/HUgfuC8WgAAE4KU.jpg","thumbnailUrl":"https://pbs.twimg.com/media/HUgfuC8WgAAE4KU.jpg","isAccessibleForFree":true,"articleSection":"Collections","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Collections","logo":"https://media.daily.dev/image/upload/s--fk_6ycEi--/f_auto,q_auto/v1780996001/logos/collections?_a=BAMAMiWQ0","url":"https://daily.dev/sources/collections"},"commentCount":6,"discussionUrl":"https://daily.dev/posts/anthropic-and-openai-sought-control-over-what-outside-ai-evaluators-can-publish-reuters-reports-jimtkttzj","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":16},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":6}],"keywords":"llm,openai,anthropic,ai-safety,ai-governance","timeRequired":"PT3M"}
{"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Collections","item":"https://daily.dev/sources/collections"},{"@type":"ListItem","position":3,"name":"Anthropic and OpenAI sought control over what outside AI evaluators can publish, Reuters reports"}]}
{"@context":"https://schema.org","@type":"WebPage","@id":"https://daily.dev/posts/anthropic-and-openai-sought-control-over-what-outside-ai-evaluators-can-publish-reuters-reports-jimtkttzj","comment":[{"@type":"Comment","text":"Anthropic and OpenAI are a joke and they are standing in the way of innovation. People need to do business with better companies and let those two fail.","datePublished":"2026-10-13T15:14:37.446Z","url":"https://daily.dev/posts/jimtkTtzJ#c-MGV5BK8Jd","author":{"@type":"Person","name":"GenXdev","url":"https://daily.dev/devgenx","image":"https://media.daily.dev/image/upload/s--PrWOWtSq--/f_auto/v1789486214/avatars/avatar_ytCVp6dQicn9P0yKfhQCI?_a=BAMAMicg0"},"interactionStatistic":{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":6}},{"@type":"Comment","text":"I’m starting to feel the alignment problem is about the AI not aligning with their own interests, not with humanity’s interests. We are yet to see a model thinking of going against humanity.","datePublished":"2026-10-14T03:14:02.799Z","url":"https://daily.dev/posts/jimtkTtzJ#c-ZpG6iQCWt","author":{"@type":"Person","name":"JP de la Torre","url":"https://daily.dev/jpdelatorre","image":"https://media.daily.dev/image/upload/s--vMLSF3-1--/f_auto/v1787723050/avatars/avatar_0fMQQvxSIIVibBvP847Xo?_a=BAMAMicg0"},"interactionStatistic":{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":6}}]}
{"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/anthropic-and-openai-sought-control-over-what-outside-ai-evaluators-can-publish-reuters-reports-jimtkttzj#faq","mainEntity":[{"@type":"Question","name":"Do Anthropic and OpenAI have the right to approve or edit reports from independent AI evaluators before publication?","acceptedAnswer":{"@type":"Answer","text":"Yes, according to Reuters reporting, both companies have sought control over evaluator publications. Anthropic says it retains comment-and-approval rights only for new evaluator firms while they learn its models, while OpenAI says it redacts confidential details from reports but allows reviewers to flag where redactions were made, as it did with a METR and Redwood Research report on a Hugging Face probing incident. daily.dev surfaces reporting like this for developers tracking how much they can trust AI safety evaluations."}}]}
```

