Thomson Reuters limits AI licensing

News licensing is getting more precise — and Thomson Reuters just made that plain. Steve Hasker said the company’s AI deals around Reuters content are limited to archive text, not a general right to ingest everything Reuters publishes. That means no live news feed, no pictures, no video, and no audio in those agreements. The point is simple: in the AI era, “Reuters content” is not one blob of rights anymore. ### What actually got limited? The limit is the package itself. Reuters sells content in separate buckets — breaking news, archives, pictures, video, audio, and partner material. Its own licensing pages make that structure pretty obvious, with archive products and live newsroom products marketed separately. So when Hasker says the AI deals cover archive text only, he is drawing a line inside an existing commercial system, not inventing a brand-new distinction. (reutersagency.com) ### Why does “archive text only” matter? Because archive text is the safest thing to license. It is finite, organized, and easier to price. Live news is different — it has immediate market value, exclusivity issues, and much higher competitive sensitivity. Pictures, video, and audio also carry their own rights chains and commercial uses. If you let an AI platform treat all of that as one training pile, you risk giving away the premium parts of the business along with the back catalog. (reutersagency.com) ### Which companies are involved? The names attached to this are Microsoft and Meta. Thomson Reuters has an official Microsoft relationship around Copilot in professional products, especially legal workflows. Meta also has a Reuters arrangement tied to Meta AI and news answers. But Hasker’s clarification matters because it pushes back on the idea that these partnerships equal full Reuters access. They do not. They are narrower than that. (reutersagency.com) ### Why make the distinction now? Because the default fight in media and AI has been over scraping first and permission later. Publishers keep arguing that platforms act as if anything crawlable is fair game. At the same time, courts are starting to let some publisher cases move forward, which raises the pressure on AI companies to get cleaner rights. In that environment, spelling out “archive text only” is both a legal boundary and a negotiating tactic. (thomsonreuters.com) ### Is this just about protecting Reuters? Not really. It also helps Thomson Reuters sell a different story about AI. The company is not mainly trying to become a wholesale content farm for chatbots. Its bigger business is workflow software for lawyers, tax professionals, and other high-stakes users. That is why its AI messaging keeps circling back to trusted content inside professional tools like CoCounsel and Microsoft 365 integrations. Narrow licensing outside the house supports that strategy. (pressgazette.co.uk) ### What does this mean for other publishers? Basically, Reuters is showing one workable middle path. Publishers do not have to choose between total refusal and total surrender. They can license old text, keep live feeds separate, and reserve multimedia rights for other products. That kind of modular dealmaking is likely to spread, especially as news companies try to avoid repeating the platform era mistake of giving distribution partners too much leverage for too little return. (thomsonreuters.com) ### So what’s the real takeaway? AI licensing is maturing from vague access into narrower rights management. Thomson Reuters is saying the valuable thing is not just content — it is control over which content, in which format, for which use. That sounds technical, but it is the whole game now. If publishers can hold that line, they keep more leverage. If they cannot, archive deals turn into back doors for much broader extraction. (reutersagency.com)