Meta just repriced your content to zero and called it streamlining

Meta pays News Corp up to $50m a year, for at least three years, to use its US and UK journalism in AI. From 30 October, Meta's revised Commercial Terms ask every business that runs a Page, buys an ad or logs into Business Suite to grant something that looks a lot like the same right, for nothing. The clause lets Meta crawl any domain "associated with" your products, pull updates periodically, create test credentials, and use what it finds to train Meta products. There is no opt-out. If you don't agree, you stop using Meta.

I expected the usual cycle of a lawyer-drafted land grab, a denial and a week of outrage. The denial arrived on schedule. Andy Stone says the publishers' reading is false, Meta says robots.txt blocks still apply, and the company calls the change a tidy-up of language scattered across older terms. What surprised me was how badly that explanation holds up when you put the old text next to the new. The Ad Creative Generative AI Terms allowed crawling of URLs you supplied for use in connection with your ads. The test-credential wording Meta pointed Ars to sits in a Seller Agreement covering e-commerce listings. Both were narrow, purpose-bound permissions. The new clause stretches them across every Meta product, with AI at Meta named elsewhere in the terms as something the content can train.

Read that as a buyer of content and it is a price change. Meta already has a rate card for this material. Beyond News Corp, it signed AI data agreements last December with USA Today, People Inc., CNN, Fox News, The Daily Caller, the Washington Examiner and Le Monde. OpenAI was reportedly offering news groups $1m to $5m a year in 2024. Google pays Reddit about $60m a year and OpenAI reportedly pays about $70m. Wells Fargo thinks Reddit's licensing income could reach roughly $550m after renegotiation, against a run rate near $130m. Sellers who could say no set those prices. A clause in self-serve terms that every advertiser must accept sets a competing price of zero, and it lands hardest on the long tail of publishers and data vendors who never got a call from Meta's partnerships team.

Meta's best defence is that it doesn't need the clause at all. In June 2025 Judge Vince Chhabria ruled in Kadrey v Meta that training Llama on copyrighted books was fair use. If training is free under copyright law, a contract adds nothing, and the line that nothing has changed would be true in substance.

I don't buy it, and the reason sits in the same ruling. Chhabria wrote that plaintiffs with evidence of market dilution, news among them, would likely win on the fourth fair-use factor. So the category where fair use looks shakiest for Meta is the one where it is paying up to $50m a year, and a contractual licence covers exactly that gap. I can't see why Meta's lawyers would draft that hedge if they thought the risk was zero. Matt Topic of Loevy & Loevy says publishers should assume the provision is enforceable even though no court has tested it, and I think that is the only sane way to run a business against it.

Then add the link cap. Meta tested limiting some Pages and professional profiles to two organic link posts a month in December 2025, and in September tied lifting the cap to a paid Meta One subscription. Publisher Pages are exempt for now. For a B2B software or data company outside that exemption, the sequence is ugly: pay Meta to post your links, and in the same account accept terms letting Meta crawl what sits behind them. Stephen Beale of The Steampunk Explorer called it "such an arrogant, in-your-face move" and is weighing quitting Facebook. Most companies won't leave, because their ad budget runs through Business Suite.

So I think commercial and procurement teams should log every platform terms update as a repricing event and cost it like a supplier raising fees. If you sell data, research or content to anyone, your licences need an order-of-precedence clause stating that the negotiated deal overrides any self-serve terms the same buyer makes you accept elsewhere, plus a fee schedule for use that arrives through that back door. If you already license to Meta, raise it at renewal, because Meta's negotiators will bring the 30 October terms into the room.

Before 30 October, open Business Suite, list every domain connected to your Pages, catalogues and ads, and check each one's robots.txt for an explicit block on Meta's crawlers. Then ask your Meta account contact for written confirmation that the commercial terms grant no training rights over those domains, and file the reply for your next price negotiation.

Prompted by Meta may use sweeping new commercial TOS to avoid millions in AI licensing fees, Ars Technica.