A Bill to Require AI Training Datasets to License Copyrighted Works
Requires commercial AI developers to license copyrighted works used for model training.
Commercial AI developers must obtain a license for any copyrighted work used in training a general-purpose AI model with over 10 billion parameters.
Developers must publish a manifest of training datasets.
Statutory damages up to $30,000 per infringed work.
Civil enforcement; private right of action.
Mechanical parts, sourced & timed
Use this as your pre-round checklist. Memorize the source citation. Time yourself to the delivery target.
- Bill / Number
- H.R. 962 — A Bill to Require AI Training Datasets to License Copyrighted Works
- Funding source
- Private right of action + civil enforcement; no new appropriation.
- Timeline
- Effective on enactment.
- Realistic — Copyright Act already structures statutory damages.
- Enforcing agency
- Private civil enforcement + Copyright Office guidance.
- Partially — Copyright Office can guide; courts ultimately decide.
- Penalty for non-compliance
- Statutory damages up to $30,000 per infringed work.
- Source citation
- U.S. Copyright Office (2024), 'Copyright and Artificial Intelligence: Part 2' — copyright.gov.
- Delivery time (read aloud)
- 1:10 (70s)
Mandatory licensing kills U.S. AI competitiveness — China and EU will train without these constraints.
The EU AI Act (2024) already imposes training-data transparency stricter than this bill. On China — they're not training on the same Western corpora at issue. And U.S. courts in NYT v. OpenAI (2024) are heading toward license requirements regardless; the bill brings legal certainty rather than imposing new constraints.