On 28 October 2026 the Oslo District Court barred Meta from enforcing new terms for professional accounts until their legality is finally determined. The Local, citing AFP, reports that the order came on an emergency basis, without Meta being heard. Ten Norwegian media groups, including NRK, Schibsted, Amedia, Polaris, Aller and NTB, brought the case. They argued the terms would let Meta collect their articles, photos and videos, paywalled material included, for AI products, and that the only way to refuse was to delete their accounts.
The case for Meta's position
Meta's side deserves a fair hearing. Platforms must be able to update commercial terms, and AI development is a legitimate use of data. Europe's AI sector will not compete if every update to terms of service becomes a legal fight. Meta also says it wants to explain the changes. PPC Land quotes Andy Stone, Meta's vice president of communications, saying the company does not plan policy changes "without a full consultation with industry groups and leaders". Meta added that the consultation could run past 20 November, in which case it would delay implementation, and that the existing terms stay in place meanwhile.
What the court actually decided
The order is narrow in what it settles. It is interim relief. The court has not ruled on whether the terms are lawful, only that they cannot take effect until that is decided. According to PPC Land, the association's announcement does not say whether the ban covers only the ten applicants or every Norwegian business using Meta's products. Meta has also not said whether it will contest the order. Anyone treating this as a precedent on AI training should wait for a ruling on the merits.
The judge's reasoning, as AFP reports it, was that media groups "risk losing control of editorial content built up over many years". That is a harm-based, interim-relief rationale and not a finding about copyright.
Why the dispute landed in contract law
The timeline explains the urgency. Euronews reports that Meta announced the commercial-terms change on 8 October, planned to roll it out on 30 October, and then moved the date to 20 November. The publishers sought an injunction on 27 October, and the court granted it a day later.
EU copyright law already has a framework for this question. The Commission's copyright legislation page lists the 2019 Directive on copyright and related rights in the Digital Single Market, which includes the text-and-data-mining exception. The AI Act builds on it. As reproduced by artificialintelligenceact.eu, Article 53(1)(c) requires providers of general-purpose AI models to put in place a policy to comply with Union copyright law, including "a reservation of rights expressed pursuant to Article 4(3) of Directive (EU) 2019/790". The Commission's page on the General-Purpose AI Code of Practice says its Copyright chapter offers providers "practical solutions to meet the AI Act's obligation" to have such a policy.
That framework assumes a rightsholder can reserve its rights. The Norwegian dispute shows the gap. A publisher who is also a platform tenant may have no practical way to reserve anything if the platform's terms grant the licence up front. Opting out then means leaving Facebook and Instagram, where the audience is. The opt-out exists in law but not in the product.
There is also a gap over robots.txt. A Meta spokesperson said the delay would be used to explain that content owners can still use robots.txt to signal preferences. PPC Land quotes Nieman Lab saying that no version of the terms has included language about respecting those preferences. A promise made in a press statement is not a term of the contract.
Why a light-touch approach should still favour the publishers here
A pro-innovation view does not mean siding with whoever has the most compute. Innovation depends on stable, predictable rules about who may use what. Three points follow.
First, consent that cannot be refused is not consent. Terms that bundle essential distribution with an AI licence and offer account deletion as the only exit are closer to compulsion than to a commercial bargain. That is a proportionate concern about contractual fairness, and it does not need new AI-specific rules.
Second, the cheapest fix is a technical one. Meta could write its robots.txt commitment into the terms, separate the AI licence from the distribution agreement, and make the AI licence optional. A narrower clause would probably have avoided the lawsuit.
Third, the market is already moving without regulators. Euronews reports that more than 340 local and national news sites have blocked internet archive crawler bots since May. Publishers are protecting themselves with technical controls, and platform terms that override those controls will push them further toward blanket blocking, which is worse for readers and for AI developers alike.
The Commission is also looking at tooling. Its copyright page lists a feasibility study, dated 13 July 2026, on an EU-level registry of text-and-data-mining opt-outs. A shared, machine-readable registry would serve both sides better than case-by-case litigation, because AI developers would know what they may use and publishers would have one place to say what they reserve.
What to watch
The next steps are Meta's consultation, whether the order is appealed or narrowed, and the hearing on the merits. If Meta revises the terms to make the AI licence optional, this episode will look like a successful market correction. If it does not, expect more national courts to be asked the same question, with different answers. For now, the lesson is that the opt-out rights European law promises have to be usable in practice, and that includes platform contracts.