Meta Removes Clause Permitting AI Training After Criticism

Meta has amended the wording of its commercial terms and removed a clause that could be interpreted as allowing content on publishers’ websites to be used to train its AI. Anthony Ha of TechCrunch AI reports.

Meta had recently updated its commercial terms, stating that it may use “any content obtained from a business account’s domains, web pages, and URLs for the provision, development, improvement, research, and learning of Meta products and services, as well as for maintaining integrity.”

Legal experts interviewed by Nieman Lab called the provision “a notable clause.” That is because it could grant permission to crawl the sites of publishers with Meta business accounts and repurpose their content for training AI models.

A Meta spokesperson maintained that the terms update did not change how the company crawls the web. The Verge later reported that the clause had been removed.

Details of the Contested Clause

The dispute centered on a sentence defining the scope of the commercial terms. While it appeared to be limited to “business accounts,” it was structured so that it could be interpreted as broadly permitting use of content from the “domains, web pages, and URLs” reachable from them.

The listed purposes included “development,” “improvement,” “research,” and “learning.” The fact that a term referring to AI model training was explicitly stated in the terms drew the attention of legal staff.

Businesses using Meta’s advertising or business products would be subject to this clause for content on their own sites. When the expert interviewed by Nieman Lab described it as “a notable clause,” they were referring to this breadth of application.

Risks Flagged by Legal Experts

The focus of the issue is not the technical behavior of crawling but the scope of contractual permission. As long as the wording of the terms can be read broadly, it leaves room for businesses to be interpreted as having “granted permission,” even if the actual crawling method is the same as before.

From publishers’ perspective, this could amount to effectively accepting, through consent to the terms, the possibility that their content will be used as training data. The lack of procedures for individual negotiation or opt-out in the terms was another factor heightening concerns.

Meta’s Initial Explanation and the Pushback

A Meta spokesperson explained that the terms update did not change the reality of crawling. The position was that it was a wording cleanup, not a change in data-collection practices.

This explanation at the same time suggests the possibility that the wording of the terms and actual practice do not match. If what was permitted as a contractual document remains ambiguous, businesses will be unable to determine the scope of their rights from the provisions. This is why legal departments spend time on review.

How the Clause Came to Be Removed and the

Official Statement

Following The Verge’s report, Meta moved to remove the clause and clarify its explanation. Its statement is as follows:

“We have listened to the feedback and want to be clear. The crawling clause we integrated applies only to advertising and business products, and we have clarified that point in the terms themselves. This does not change how Meta uses web crawlers.”

The statement again indicates that the clause applies only to advertising and business products. However, it does not explain why the removed wording was initially integrated in such broad language.

The Situation Around AI Training and

Copyright Litigation

According to TechCrunch AI, Meta won a high-profile copyright infringement lawsuit last year. Meanwhile, other lawsuits against AI companies, including Meta, are ongoing.

Under such judicial circumstances, a single sentence in the terms can affect arguments in litigation. This is because whether a company can show that it had “obtained permission through its terms” for acquiring training data can become a pillar of its defense. Conversely, if the wording of the terms is too broad, it can invite new points of dispute.

The removal can be seen as a measure to maintain consistency between arguments in court and external explanations. It is an example showing that we have entered an era in which a single line of wording directly affects business risk, as companies treat their terms as public documents.

Impact on Publishers and Developers

The editorial team sees this case as an opportunity to change how terms of use are read. Until now, terms have tended to be treated as a checkpoint to pass through via consent screens. But as debate over training data for generative AI continues, a single word in a provision can determine the scope of business permission.

Among publishers, there may be a growing movement to review crawl permissions for their sites on both technical and contractual fronts. This is because settings in robots.txt alone cannot manage the scope of contractual permission.

For developers, it becomes necessary to continuously monitor revisions to the terms of platforms they use in their work. In particular, when incorporating advertising or business products into their own services, it is essential to check whether the scope of the terms extends to content on their own sites.

What the Terms Change Signals Going Forward

What is notable is that what was removed was the clause itself, not the practice of crawling. Meta says it has not changed its practices. The criticism nonetheless arose because the meaning of the permission indicated by the terms cannot be separated from practice.

The trend for AI companies to specify in their terms the routes by which they procure training data is expected to continue. Publishers, too, will move toward clarifying in contracts the scope of permission and whether there is compensation. As a result, negotiation costs for procuring training data will rise.

On the other hand, efforts to make the wording of terms more plain may undermine legal comprehensiveness. The industry has yet to present a common solution to this trade-off.

Editorial Opinion

In the short term, the revision of the terms wording may prompt other platforms to review their own terms. We expect a broadening movement among legal departments to re-examine interpretations of crawl clauses. The fact that the process from announcement to removal moved quickly shows that the visibility of terms was recognized as a business risk.

From a one- to three-year perspective, the trend toward explicitly stating in contracts the scope of permission for data used in AI training will strengthen. We assess that publishers will put in place both crawl settings and contractual terms. The cost of procuring training data will rise, and smaller businesses will be more susceptible to the impact.

The issue is why a wording change that does not alter the reality of crawling caused misunderstanding. There is still no industry-wide standard for whether to prioritize plainness or legal comprehensiveness. A perspective of reviewing the permission conditions for one’s own site is required.

This article is based on TechCrunch AI (All Rights Reserved). It relies on fair quotation under Article 32 of the Copyright Act of Japan.

References

• “Meta removes language that raised scraping and AI training concerns”, by Anthony Ha — TechCrunch AI, 2026-10-17T21:31:32.000Z (ARR)
• Source URL: https://techcrunch.com/2026/10/17/meta-removes-language-that-raised-scraping-and-ai-training-concerns/