Britannica, Merriam-Webster sue OpenAI over training data

This case sharpens the legal pressure on foundation-model training: courts are being asked to define where “fair use” ends and commercial substitution begins.

• –The complaint targets both model training and retrieval-style use, not just isolated output examples.
• –Dictionary and encyclopedia content are high-signal reference data, so restrictions here could materially affect model quality and grounding pipelines.
• –If plaintiffs win meaningful relief, licensing costs and provenance requirements for AI developers could rise fast.
• –Even without a final ruling, the filing adds momentum to broader publisher litigation and may push more preemptive content-licensing deals.