Executive Summary

Key figures

Source: the CMA announcement and the case page, both dated October 30, 2026.

The case is filed as "Suspected abuse of a dominant position in relation to Meta's commercial terms," and the parties under investigation are Meta Platforms, Inc., Facebook UK Limited, Facebook Ireland Limited and other entities in Meta's corporate group with a UK connection. The CMA has set out two questions. Whether Meta holds a dominant position in one or more relevant markets in the UK, and, if it does, whether its existing commercial terms, or revisions to those terms, amount to an abuse of that position under Chapter II of the Competition Act 1998.

The target is not the document as a whole but the passage on collecting content. In the CMA's description, the terms could allow Meta to collect content from the websites of businesses and publishers that use its services and then use it to develop and provide its own products and services, its AI features included, without those businesses getting a choice on if or how their content is used. Requiring news publishers and other content businesses to accept that, the regulator wrote, could amount to an unlawful imposition of unfair trading conditions for thousands of UK businesses. The summary line on the announcement uses a blunter word and calls the terms exploitative.

"Digital platforms play an important role in helping UK businesses reach millions of customers around the world. But where a company holds a dominant position, it must not use that position to impose unfair terms."

Sarah Cardell, Chief Executive of the CMA, in the announcement published on October 30, 2026.

The CMA did not confine the affected group to news publishers. It wrote of thousands of UK businesses that rely on Meta's services to reach customers and readers, and advertisers sit inside that number. For now, though, this is only the opening of a case. The case page states that no assumptions should be made at this stage about whether competition law has been broken, and the published timetable holds a single row, rapid initial information gathering and engagement with Meta and third parties from October 2026 to January 2027, with a note that the dates are indicative.

These terms moved three times in a single month. When the revised version first went up, it carried a sentence letting Meta use automated software to gather content from a business's domains and use it to develop and train its own products. As the backlash grew, Meta took the word train out of the US English edition, and legal experts pointed out that the surviving word develop is broad enough to hold training inside it. The piece Pebblous published on October 19 dealt with how unevenly the deletion had landed. The UK English edition still carried train, and the Korean edition still carried a word in that slot that reads back as education.

Then the date started sliding. Meta had announced October 30 as the effective date, moved it to November 20 after engagement with the CMA, and said on October 28 that it was pausing the rollout entirely to consult industry groups. A story would naturally end there. Two days later the investigation opened. That sequence is laid out below.

A pause fails to stop the case because of what competition law looks at. The second of the CMA's two questions reaches both the existing commercial terms and any revisions to them, which puts the version already in force inside the perimeter alongside the one whose start date slipped. When the investigation is about where the power to set a condition came from, setting that condition aside for a while is no answer.

Fights over AI training data have mostly been held in copyright courts. A copyright court asks who wrote the text, whether the party that took it copied it, and whether the use was fair. This case puts a different question to the same facts. It looks past the act of taking the content and examines the design that made accepting the terms the price of staying on the platform.

What Chapter II prohibits is the abuse of a dominant position rather than the holding of one. The investigation therefore stands on two steps. Meta's dominance in the relevant markets has to be established first, and only then can the terms be judged an abuse. Put the other way around, the same terms offered by a small company would fall outside what this law can reach. The issue is less the content of the condition than what a business is left with after refusing it.

Trade bodies state that point most plainly. Theo Bamber, chief executive of the News Media Association, welcomed the investigation and summed up what publishers were offered: "Meta has given publishers an invidious choice of giving away their content for free to power its AI products or leaving a platform which millions of Britons use to get their news." When only one of the available options is survivable, whether the resulting consent counts as consent is the substance of this case.

Meta describes the arrangement differently. It has said the crawling terms applied only to its ads and business products and that content owners could still use robots.txt to signal their preferences, and it has said it will respect those signals. That promise sits outside the terms rather than inside them. What the document grants is permission to collect and use the content, while the undertaking to respect a signal lives in statements to reporters. A condition written into an agreement can be enforced against the party that breaks it, and an undertaking made beside the agreement can be withdrawn. In an investigation about whether anyone had a box to check, that distinction matters.

There is also a question of powers behind the CMA's choice of door. The Digital Markets, Competition and Consumers Act 2024 created a designation called Strategic Market Status, which lets the regulator attach tailored conduct rules to a designated firm. Google has been designated in search, and Microsoft is in the process over business software. Meta carries no such designation. The CMA says it keeps potential candidates under active review and that the next set of proposals goes to the CMA Board in spring 2027, so the tool that would help here is not available yet. The road left open was the Competition Act 1998. Reaching a conclusion on it takes a long time, and a confirmed infringement can carry a fine of up to ten percent of worldwide turnover.

Britain is one of several places where these terms have run into trouble. Ten news companies went to court in Norway and an Oslo court temporarily blocked the terms on October 28. France's competition authority imposed interim measures on Meta in July over talks on paying publishers for their news content. Complaints have been filed with the competition authorities in Denmark and Sweden, and complaints to the same effect have reached the European Commission. What has been made public so far sits in the table below.

Reading the five rows as one block would flatten them, though, because the procedures differ in kind and some of them bite today while others await a ruling. Norway's is a prohibition already in force. The ten companies the court sided with include the public broadcaster NRK, TV 2 and the media group Schibsted, and the terms are blocked there until a court has reviewed whether they are legal. The UK is at the stage of opening a case, which carries no power to stop the terms now and a fine later if an infringement is confirmed. Denmark and Sweden are known to have reached the filing of complaints. In the European Union, the European Publishers Council asked the Commission to halt the terms before November 20, the date Meta had briefly moved to, and framed the request under both the Digital Markets Act and competition law, while Movement for an Open Web, which complained to the CMA, filed with the Commission as well.

Sources: Press Gazette and PPC Land, as reported on October 30, 2026. The UK row follows the CMA case page.

The French row covers a slightly different matter. The interim measures that France's competition authority imposed in July were an order to negotiate with French publishers over related rights, the payments owed when their news content is reused, and they had nothing to do with the crawling terms at issue here. Meta appealed and lost in Paris on October 29. The subject is different and the question still overlaps. Both come down to who sets the terms when a publisher's work ends up in a platform's product.

Meta's position is that there are no grounds for a competition case. A spokesperson said the company had already committed to consulting publishers and other stakeholders before changing its terms and welcomed the chance to explain its approach to the CMA. One side treats the commitment as sufficient and the other is looking at the power to set a condition in the first place, which is a distance unlikely to close while the investigation runs.

Take the word publisher out of the story and the shape left behind is a familiar one. Using a service means agreeing to its terms, and somewhere in those terms is a sentence letting the other company put your data to work on its models. Where the consent screen has no box for declining, that clause arrives attached to the decision to use the service at all. A fair share of the SaaS and API agreements running inside any company already rest on that shape, and the only difference here is the size of the counterparty.

So the investigation leaves three questions with anyone responsible for data.

• • Does the platform agreement you work under contain a data use clause. Uploaded material, logs and prompts all need checking sentence by sentence against how far the other company's product development is said to reach. Broad words such as improve and develop can hold training inside them, which is exactly what the experts put their finger on in this case
• • Is there a real way to decline. Whether a switch exists in the settings, whether that switch is tied to a pricing tier, and what else goes dark when you flip it all belong to the same question. Where declining means sending an email to support, that is an application process rather than an option
• • Do the language and region editions say the same thing. After Meta cut the word train from the US English edition, the UK English edition kept it and the Korean edition kept a word that reads back as education. Judging from the head office English edition alone leaves you unaware of the sentence you actually agreed to

When we talk about AI-Ready Data, most of the work is getting data into a state a model can use. Schemas in order, quality metrics attached, provenance recorded. This investigation adds one more question to the provenance part. What belongs in the record is not only where the training data came from but whether the party it came from had anywhere to say no. Absence of consent is among the hardest things to reconstruct later from the model end.

Thanks for reading this far. The facts here come from the CMA announcement and case page, and the international picture and Meta's position come from reporting by Press Gazette, PPC Land and others. The investigation is at the initial information gathering stage and a conclusion is some way off. We plan to return to it if interim measures or a provisional finding arrive. We would like to hear how you check the data use clauses in the platform agreements you work under.

Pebblous Data Communication Team
November 1, 2026

Official sources

• 1.Competition and Markets Authority. (2026). "Suspected abuse of a dominant position in relation to Meta's commercial terms." GOV.UK.
• 2.Competition and Markets Authority. (2026). "CMA opens investigation into Meta's commercial terms." GOV.UK.

Reporting

• 3.Press Gazette. (2026). "UK competition watchdog opens investigation into Meta scraping terms."
• 4.PPC Land. (2026). "Meta faces UK probe over planned terms to use publishers' content for AI."