Meta strikes multi-year deal with Reuters to bring real-time news to Meta AI

Meta just brought real-time news into its AI chatbot through a multi-year licensing deal with Reuters, Axios first reported.

Rolling out immediately to U.S. users, the feature allows Meta AI to pull from Reuters reporting whenever someone asks about news or current events. Every response will cite the wire service and provide a direct link to the original article. Meta and Reuters both confirmed the pact, though neither disclosed the financial terms or whether Reuters copy can be used to train future Llama models.

Real-time reporting inside WhatsApp and Instagram

Meta AI is already embedded into the search bars and chat threads of Facebook, Instagram, WhatsApp, and Messenger, as well as the Ray-Ban Meta smart glasses.

Until now, Meta positioned the bot mostly around creative writing, image generation, and helping users with homework or shopping. When questions turned to breaking news, chatbots famously struggled with hallucinations. Under this deal, Reuters gets compensated for access to its journalism, while Meta gets a reputable, fact-checked pipeline for real-time information.

Interestingly, Reuters has worked with Meta as a fact-checking partner since 2020, but it sat out Meta’s previous, multi-million-dollar publisher deals from the 2019 "News Tab" era.

Why it matters

Meta spent the last few years quietly scrubbing political content and news from its feeds, largely to dodge regulatory headaches. But AI chatbots change the equation entirely: when an assistant sits inside your messaging app, you expect it to know what happened ten minutes ago.

We think picking Reuters as a first partner is a calculated, low-risk move. Unlike opinion-heavy outlets or magazines, a global wire service gives Meta clean, non-partisan facts and quick updates. That minimizes political backlash, especially around election cycles.

It also highlights how the AI arms race is forcing tech giants to pay up. OpenAI has already signed deals with News Corp, the Financial Times, and Vox Media, while startups like Perplexity face lawsuits from publishers for scraping without permission.

The unanswered question here is training data. OpenAI’s deal with the FT explicitly allowed article usage for model training. If Meta is only licensing Reuters for real-time retrieval (RAG) rather than training Llama, it solves the immediate answer problem, but won't fundamentally improve the underlying model's built-in worldview.