Training Data Copyright Litigation 2025
Tags: Regulatory, Case Studies
TL;DR
- In April 2025, the court denied OpenAI’s motion to dismiss the NYT’s copyright claims tied to training on NYT content.
- The case underscores unresolved risks around training data, fair use, and outputs replicating copyrighted works.
Why it matters for HK marketers: IP exposure persists when vendors train models on publisher content; expect tighter licenses and indemnity demands.
What the court did
- Motion to dismiss denied: The NYT presented sufficient claims to proceed, keeping core copyright questions alive.
- Commercial context: Publishers are pursuing litigation and licensing deals to monetize training use of their content.
Adjacent rulings to watch
- Courts are also scrutinizing training on proprietary databases (e.g., rulings finding certain headnotes copyrightable and temporary copies not fair use in training contexts).
So what for marketers
Prefer vendors with clear content licenses; include warranties against training on unlicensed publisher data and require output filtering for verbatim reproductions.
← Back to Knowledge Base