2026.10 – Meta Commercial Terms: Scraping Business Websites Under EU Law – Interview

On October XX, 2026, Pierre-Xavier Chomiac de Sas, a Paris-based avocat and certified specialist in digital law, was interviewed about Meta’s new commercial terms, which take effect on October 30, 2026. PCS Avocat sets out its analysis below.

Published and then revised within a few days, these terms show Meta’s intent to let its bots crawl the websites “associated with” its users’ business accounts, well beyond the content posted on its platforms: Facebook, WhatsApp, Instagram, Messenger and others.

Highly favorable to Meta, the terms raise doubts about their enforceability and about how recent European legislation protects the businesses concerned: the AI Act and artificial intelligence law (in French), the Digital Markets Act, the GDPR and press publishers’ related rights.

Meta Commercial Terms – A Two-Step Revision

In early October 2026, Meta notified business account holders of an update to its commercial terms. The update covers accounts that manage pages or groups, run ads, send marketing messages and, more broadly, use the business features Meta makes available to them.

On October 15, Nieman Lab, the journalism publication of Harvard’s Nieman Foundation, published the views of three US legal experts warning about the consequences of this new contractual framework. In their reading, the terms would allow Meta to collect content from news websites to train its AI products.

On October 16, Meta published a revised version. The word “train” was removed from the permitted uses, which are now limited to advertising and business products. Several publishers’ organizations, including the European Publishers Council, announced they would bring the matter before the European Commission.

Scraping Clause – A Scope Beyond Meta’s Platforms

The scraping clause in Meta’s terms does not concern content posted on Facebook or Instagram. It targets businesses’ own external websites.

“Associated” websites. The terms authorize Meta, or any entity acting on its behalf, to crawl the domains, web pages and URLs “associated with” the business’s products or services. Nothing limits this wording to a product page or a catalog. Read broadly, it covers a business’s entire website, including a news publisher’s.

Shared links. The terms also cover URLs made available through the business’s “or users'” use of Meta products. Here again, a broad reading would extend to any hyperlink shared by third parties on the platform, including links to the website of a business that has no business account at all.

A generalization. Until now, this crawling right appeared only in the terms governing Meta’s generative AI advertising tools, where it served to optimize ads. Meta describes the update as a mere consolidation. Yet the change in scale is clear: simply using Meta’s business services now amounts to consent. In the initial French version of the terms, the verb “train” applied to content on external websites, while it did not appear for content posted on the platforms.

Removing the word “train”. The revised version limits use to “advertising and business products.” The new wording clarifies the purpose, but it does not appear to narrow the scope of the data collected. A generative AI tool for advertisers is still a business product, and developing it requires training it.

Test Credentials and Paywalls – Access Granted by Contract

The terms allow Meta to “create test credentials” to access associated domains. They contain no exclusion for paid or restricted areas. Asked about this by Nieman Lab, Meta did not respond.

The risk is not primarily technical. It is legal: access consented to by contract neutralizes several grounds for legal action.

Access to an information system. Article 323-1 of the French Criminal Code (Code pénal) punishes fraudulent access to, or remaining in, an automated data processing system. Access authorized by terms the publisher has accepted is not fraudulent. This cornerstone of French digital criminal law (in French) becomes difficult to invoke.

Technological protection measures. Articles L. 331-5 et seq. of the French Intellectual Property Code (CPI) protect the technological protection measures (TPMs) that prevent unauthorized uses of a work. A credential accepted by contract does not circumvent any measure: it unlocks it.

Text and data mining. The text and data mining (TDM) exception in Article L. 122-5-3 of the IP Code, which implements Article 4 of Directive (EU) 2019/790, applies to lawfully accessible content. Meta could rely on a contractual access authorization to establish that lawful access, even behind a paywall. The same reasoning could be raised in any EU Member State, since the exception stems from the same directive.

News Websites – Rights Granted Without Negotiation

News groups are the first to be affected by this change, given the sheer volume of content published on their websites. Almost every newsroom has at least one Meta business account. They are therefore exposed to large-scale collection of their articles for Meta’s sole benefit, with no negotiation over the value or licensing of their related rights.

The only way to refuse would be to stop using Meta products for business purposes, which no publisher can realistically do.

A weakened licensing market. In December 2025, Meta signed agreements with several outlets, including CNN, Fox News, USA Today, People Inc. and Le Monde, to integrate their news content into Meta AI, with links to their articles. Similar agreements followed with Le Figaro, Prisa and Süddeutsche Zeitung. A blanket collection clause undermines publishers’ position in these negotiations.

The Google precedent. In France, the press publishers’ related right (droit voisin), created by Act No. 2019-775 of July 24, 2019, which implements Article 15 of Directive (EU) 2019/790, requires prior authorization from publishers for online reuse of their content. The French Competition Authority (Autorité de la concurrence) has fined Google twice on this basis: €500 million in 2021, then €250 million in Decision No. 24-D-03 of March 15, 2024 (in French). The Authority found that Bard had been trained on news content without publishers being informed. It also faulted Google for offering no way to opt out of this use without affecting how their content was displayed in Google’s other services. Tying access to a service to the use of content by AI is precisely the mechanism that Meta’s terms now generalize.

Legality of Meta’s Commercial Terms – The Safeguards of EU Law

As is standard for US platforms, Meta’s commercial terms specify the governing law and give jurisdiction to the courts of California or Ireland. These contractual provisions may, however, run up against EU directives and regulations that limit their reach.

Digital Markets Act. Meta has been designated a gatekeeper for Facebook and Instagram under Regulation (EU) 2022/1925 (DMA). Article 6(12) requires fair, reasonable and non-discriminatory conditions of access to online social networks, along with an alternative dispute resolution (ADR) mechanism. Meta’s terms themselves reserve these rights in their section on European disputes. Making business access conditional on a license to crawl the business’s own website raises a direct question of fairness.

P2B Regulation. Regulation (EU) 2019/1150 governs relations between online platforms and their business users. It requires clear and intelligible terms and advance notice of any change. A collection authorization of this breadth, embedded in standard terms, sits poorly with this transparency requirement.

Text and data mining opt-out. Under EU law, an opt-out from text and data mining has legal effect. In France, it may be expressed by any means, including machine-readable means (Article R. 122-28 of the IP Code). Article 53(1)(c) of Regulation (EU) 2024/1689 (AI Act) also requires providers of general-purpose AI models to identify and respect these reservations. This sets the EU apart from the United States, where a robots.txt file is widely seen as a mere preference. In this context, a general click-through clause does not appear capable of lifting an express reservation of rights.

GDPR. The websites collected contain personal data: journalists’ names, reader comments, contact details. Large-scale collection by Meta requires a legal basis and notice to the individuals concerned, which the latest versions of the commercial terms do not specifically address. Personal data protection (in French) requirements apply to Meta as to any data controller.

Rights Warranty – Risk Shifted to the Business

The commercial terms require the business to warrant that it holds all the rights needed to grant the licenses provided “under the Commercial Terms.” This warranty would appear to cover the collection clause.

The consequences are significant, because a business website rarely contains only content the business fully owns: agency photos, articles by freelance journalists, licensed content. Meta’s terms nonetheless require the business to vouch for all of it, which exposes the business to liability if a dispute arises.

By contrast, the terms cap Meta’s liability at $100 or the amounts paid over the previous twelve months. The validity of such a clause under EU law is open to question. It nonetheless makes it harder for rights holders to bring a copyright infringement action (in French) against Meta if their content is used to train its tools.

Platform Terms and AI – Digital Law Attorney in France

A growing number of platforms now include specific clauses on AI tools and on the conditions under which they allow themselves to collect data for training. Governing these practices requires reconciling all applicable national and EU regulations. For businesses operating in France, this means combining French copyright and press law with the EU’s AI, platform and data protection rules.

PCS Avocat advises businesses on these choices, in both advisory and pre-litigation matters, as part of its artificial intelligence law (in French) practice.

• News publishers and online media: review of platform terms, drafting of text and data mining opt-outs, negotiation strategy for AI licensing deals.
• Businesses with Meta business accounts: mapping of exposed content, updates to legal notices, website terms of use and robots.txt files.
• Agencies and content creators: verification of the chain of title before granting any license to a platform.
• AI providers and integrators: AI Act copyright compliance policies, securing training datasets.
• Rights holders: formal notices, copyright infringement actions, complaints to regulators.

Further Reading

• Artificial intelligence law and AI Act compliance (in French)
• Enforcing intellectual property rights (in French)
• All the firm’s publications and speaking engagements (in French)
• Source: Nieman Lab, analysis of the revised terms (October 17, 2026)

Does your business have a Meta business account? Contact the firm to review the content you may be exposing.