Publishers Launch Meridian for AI Scholarly Access
The announced service would let AI tools cite peer-reviewed books and journals through library subscriptions, with a limited test run planned for this fall.
Updated on Nov. 3, 2026 in Artificial Intelligence
Live Poll
Should AI companies have to pay publishers when their tools use books and articles?
Cambridge University Press, Johns Hopkins University Press, and Princeton University Press founded Meridian, an AI licensing service run by Cashmere, to give students and researchers authorized access to scholarly books and journals. The service is announced rather than shipped: a limited test run with a group of libraries and institutions begins later this fall.
Why it matters
Publishers and libraries are pursuing shared rules for AI access to scholarship before outside tech platforms set those rules. Meridian routes AI tools to peer-reviewed content through a single publisher-controlled connection rather than separate deals with each AI company.
The only quantified figure is the $5 million raised by Cashmere, which was not compared against a benchmark in the source. Meridian's performance, usage volumes, and pricing have not been reported, and the platform has not yet been tested with users.
The players
Cambridge University Press
A university publisher whose scholarly books and journals are among the founding catalogs of the Meridian licensing service.
Johns Hopkins University Press
The university press of Johns Hopkins University, one of the founding publishers of the Meridian licensing service.
Princeton University Press
The university press of Princeton University, one of the founding publishers of the Meridian licensing service.
Cashmere
A Salt Lake City company that built the technology behind Wiley's licensing deal with Perplexity and now runs Meridian.
The details
When a researcher asks an AI tool a question, Meridian lets that tool pull its answer from peer-reviewed content and cite the source. Meridian sends the reader through their library's existing subscription to the final published version, and content accessed through it cannot be used to train AI models or be resold to other companies. Publishers build one connection to Meridian instead of a separate connection for each AI company, which uses the Model Context Protocol, an open standard developed by Anthropic.
Timeline
Later this fall: a limited test run with libraries and institutions begins.
This month: founding publishers load their catalogs onto Meridian.
Earlier this year: Cashmere raised $5 million from investors led by Reach Capital.
At least two years: Cashmere commits to running Meridian during its evaluation period.
The Tech Race
Meridian's approach of one publisher-controlled connection, built on the Model Context Protocol, positions scholarly publishers to negotiate AI access collectively rather than deal by deal. Cashmere's earlier work on Wiley's licensing deal with Perplexity shows the company has already built a comparable AI licensing pipeline.
Students and researchers at libraries that join the test run could eventually have AI tools cite peer-reviewed books and journals they already have access to through their libraries. Availability beyond the test run, pricing, and the timeline for wider use have not been announced.
The takeaway
Watch whether the fall test run with libraries and institutions goes ahead as planned, since it will show whether publisher-controlled AI licensing can work at scale. Also watch whether the council of publishers incorporates as an independent nonprofit to become Meridian's board.
What happens next
Founding publishers will load their catalogs onto Meridian this month, and a limited test run with libraries and institutions begins later this fall.
Further reading
For broader context on how AI tools are reshaping access to research, explore our Artificial Intelligence coverage.
Live Poll
Should AI companies have to pay publishers when their tools use books and articles?
