Britannica, Merriam-Webster sue OpenAI over training data

This is a direct legal attack on the “train first, litigate later” playbook, and it could pressure AI labs toward formal licensing at scale.

• –The case targets both model training and retrieval/output behavior, not just one narrow infringement theory.
• –Plaintiffs argue AI answers can substitute for visits to original sites, putting ad and subscription revenue at risk.
• –Trademark-style claims around hallucinated or incomplete attributed outputs widen the risk beyond pure copyright.
• –If this survives early motions, developers should expect stricter provenance, filtering, and publisher-deal requirements across AI products.