Reuters opens 39 years of news archives to AI buyers on Snowflake

Reuters is making nearly four decades of its journalism available through Snowflake Marketplace, an online marketplace for enterprise data and software, giving enterprises another way to bring licensed news into artificial intelligence tools.

The company announced the listing on Snowflake Marketplace Wednesday in a PR blog post. Reuters said the collection includes journalism dating to 1987 in five languages and can be used in AI applications that track developing events, analyze companies and industries over time and surface market or geopolitical signals.

For Reuters, the announcement is also about how that access happens.

Erika Young, the company’s strategic business development director, said the integration “allows us to participate in the AI ecosystem under proper licensing terms that uphold our intellectual property rights.”

Reuters says the marketplace gives enterprises a controlled way to access and integrate its material under license, though access is not necessarily available to every Snowflake customer.

Snowflake Marketplace describes its model as “no-copy” access, meaning customers can work with data where it is hosted rather than make separate copies. It also supports on-platform purchasing. Snowflake says that setup can lower integration costs and streamline procurement, but neither company said how long a Reuters licensing agreement would take.

The listing comes as Snowflake has been building a broader business around licensed publisher content. The Media Copilot previously reported that Snowflake’s Cortex Knowledge Extensions included 17 publishers, among them The Washington Post, The Associated Press, People Inc. and USA Today Network. Snowflake told Digiday that some publishers had signed six-figure licensing deals, structured as flat-fee or usage-based agreements.

Reuters also arrives on the marketplace as its parent company invests in its own AI systems. Thomson Reuters said this week that it spent about $40 million building a proprietary large language model called Thomson. The company said it began with an open-source foundation and trained the model on material from Westlaw, Practical Law, Checkpoint and Reuters.

Reuters has separately changed how existing news-agency customers can use its journalism with AI. In July, it launched a Model Context Protocol server that allows customers’ AI agents to search, retrieve and download Reuters content they already subscribe to. The Snowflake listing is another way Reuters content can enter AI workflows, but the company did not describe the two products as parts of a coordinated strategy.

Important commercial details remain unclear. Reuters did not disclose pricing, revenue sharing or customer numbers. It also did not say whether the Snowflake license permits model training or limits customers to retrieval, analysis and other applications.

Other companies are testing marketplace approaches. Microsoft launched its Publisher Content Marketplace this year, saying publishers can set terms for AI training and receive payment based on usage. The products are not identical, but both put a marketplace between content owners and companies seeking licensed material.

There is also a longer-term question about what happens after journalism enters AI systems. Reuters Institute researcher Nic Newman has warned about AI turning news into “liquid content”, with information repackaged into formats tailored to individual users.

Reuters’ announcement does not say how its journalism will be displayed, credited or transformed inside customers’ products. What it does make clear is that Reuters wants more AI use of its archive to happen through licensed channels — and Snowflake is now one of them.