Meta Rewrites Scraping Terms After Publishers Raise AI Training Fears

Written By James Taylor
Could a Business Account Give Meta Rights to Your Website? The Wording That Sparked a Row

Meta Removes Website-Scraping Language After Publisher Concerns

The Fight Over Crawling and Consent

The dispute centres on whether using a social platform for business could also authorise the collection and reuse of material published elsewhere.

Meta has revised controversial commercial-terms wording after publishers raised concerns that it could authorise the scraping of their websites for artificial intelligence training. The change removes a sentence describing broad uses of collected website content, including training, ahead of an update scheduled for 30 October 2026.

Meta says the permissions concern advertising and business products, do not alter its crawling practices and grant no new rights overriding website owners’ stated preferences. That clarification is narrower than a promise to stop all scraping. The dispute concerns the reach of a particular contract and what business users could be understood to have accepted.

The Permission That Alarmed Publishers

The earlier wording described automated access to websites associated with a business’s products or services. It also extended to material made available through the use of Meta products, raising concerns about how widely that connection could be interpreted.

The proposed language went beyond displaying something a company had deliberately uploaded to Facebook or Instagram. It contemplated collecting material from websites outside those platforms and using it in developing Meta products.

That distinction matters. A publisher might use a social account to distribute links while keeping the underlying journalism on its own website. Permission to operate the account and permission to reuse that journalism are different decisions.

Legal specialists reviewing the earlier text warned that its breadth could support an argument for AI training rights. Those concerns were interpretations of the proposed wording, rather than a finding that Meta had already used it to collect news content.

What Meta Says the Revision Means

Meta’s latest explanation confines the consolidated crawling terms to its advertising and business products. The company also says website owners can continue to communicate access preferences through robots.txt, the file used to give instructions to automated crawlers.

The sentence removed from the commercial-terms page had listed purposes for using collected content. Its deletion matters because an express reference to training was central to the alarm. It does not, by itself, establish that every form of automated website access has been removed from Meta’s contracts.

There is also an important distinction between changing contractual language and changing a technical operation. Meta says its crawling behaviour has not changed. The revision should therefore be understood as a response to the disputed wording, rather than evidence that an identified news-scraping programme has been shut down.

Why Publishers Objected to the Link Between Access and Consent

The European Publishers Council challenged the proposed terms on 16 October. Its executive director, Angela Mills Wade, argued that maintaining a business account or having a reader share an article should not become blanket permission to exploit a publisher’s work.

The council also questioned the practical freedom publishers have to reject expansive conditions when they depend on major platforms to reach audiences. It called for scrutiny across the European Commission’s competition, digital and copyright responsibilities.

That is an objection about bargaining power as well as content. A publisher can choose what to post, yet still face difficult choices about the conditions attached to the distribution service. The wider effect of AI on institutions and power depends partly on who gets to set those conditions.

The council’s intervention establishes its position. It does not establish that the Commission has opened proceedings or decided that Meta’s terms are unlawful.

Crawling, Training and Serving an Answer Are Different Uses

The controversy also illustrates how loosely the word “AI” can be used. A system might fetch a page to retrieve current information, build an index to locate material, or use a collection of documents during model development. Those operations raise different questions about access and reuse.

How AI learns from training data helps explain the distinction. Training shapes a model’s learned parameters. Retrieving a page supplies information to a system at the time it performs a task. Evidence of retrieval alone does not prove that the same page became training material.

For publishers, a useful explanation therefore needs to identify both the material being accessed and its intended use. A general assurance about crawling cannot answer every question about storage, training or redistribution. Equally, a broad clause cannot establish which technical activities actually occurred.

The Next Test Is the Terms That Take Effect

The approaching 30 October implementation date gives the dispute a concrete point of reference. Business users need to assess the revised language, rather than rely on screenshots or descriptions of the version that provoked the original complaints.

The remaining question is how clearly the final contract defines its scope: which websites it covers, what content may be collected and what uses follow. Meta has explained what it says it intended. The operative wording will determine how plainly that intention is expressed to the businesses being asked to accept it.

Sources

Next Reads

James Taylor
Next

AI Can Move Your Money. Who Takes Responsibility When It Fails?