Meta expands access to content collection for training its models

Meta plans to change its commercial terms on October 30, 2026. Companies using its tools could see their content used to train or improve the group’s AI models.

Starting October 30, 2026, Meta plans to allow the collection of content published by its professional clients to train its AI models or improve its products. This measure affects companies using its advertising tools, pages, groups, or business messaging. It raises a sensitive question for digital players: how far can a platform provider reuse what its clients publish on its platform?

Des conditions commerciales qui changent la donne

According to the updated commercial text, Meta reserves the right to use automated software — “web scrapers, bots, robots, spiders, crawlers, and user agents” — to access domains and pages related to its clients’ products or services. The company also states it can “index, cache, analyze, collect, retrieve, or crawl” this content, then use it to “provide, develop, improve, research, train, and maintain the integrity” of its products and services.

The scope is broad. Meta defines its professional clients as users of ads, marketing messages, product catalogs, pages, business-use groups, or tools like Ads Manager and Meta Business Suite. The group counts at least one million companies using its AI agents on WhatsApp and Messenger, not including other services, according to the cited source. For many brands, commercial life already passes through Facebook and Instagram; now, the legal line becomes more sensitive.

Pourquoi les éditeurs et les entreprises s’inquiètent

Danielle Coffey, CEO of the News/Media Alliance, denounced the measure as “outrageous and probably illegal.” In her statement, she considers that Meta conditions access to its audience on accepting content collection on companies’ sites, which she calls “an appalling, bad-faith policy, and a clear abuse of the company’s market power.” The reaction is far from anecdotal: almost all news publishers have at least one professional account with Meta, according to Nieman Lab.

Eric Goldman, law professor at Santa Clara University School of Law, sees both a procedural and a substantive problem. Formally, he recalls that a change in terms must allow a real possibility to refuse. Substantively, he points out that the text does not explicitly say Meta will respect robots.txt instructions, those files that tell indexing robots what a site allows or not. “The law doesn't necessarily compel anyone to respect robots.txt files,” he explained to the Register. Atmosphere.

Un test de confiance pour Meta

The issue is part of a heavy past. Goldman recalls controversies linked to Cambridge Analytica and the agreement with 52 US state attorneys general on social media addiction, which have damaged trust around Meta over its 22 years of data collection. In this context, the most unfavorable reading of the new clauses becomes easy for the companies concerned, especially if they store intellectual property or sensitive catalogs there.

Meta did not respond to the request for comment cited by the source. At this stage, the question is therefore not only technical or contractual; it is commercial, legal, and reputational. For companies already feeding platforms with content, the message is clear: they will need to reread the terms of use before clicking too quickly.

By the numbers

  • October 30, 2026 — effective date of the new commercial terms, according to the source.
  • At least 1 million companies — use Meta’s AI agents on WhatsApp and Messenger, according to the source.
  • 52 states — attorneys general involved in the addiction agreement, according to the source.
  • 22 years — of data collection history attributed to Meta in the cited source.

To read