Writers vs. Databricks: Deposition Dispute Puts AI Training Practices Under Scrutiny
Writers suing Databricks for allegedly using their copyrighted works to train artificial intelligence models have asked a California federal judge to stop defense attorneys from coaching witnesses during depositions. Defense counsel pushed back, arguing the court should bar plaintiffs from asking "personally invasive and harassing" questions.
It's a procedural fight, but it signals something bigger: how far each side is willing to go to test the facts behind AI training. For working writers, the outcome could influence how consent, licensing, and data provenance are treated in future disputes.
Why this matters for writers
These cases are shaping how unlicensed training data is handled and what accountability looks like. If the court tightens behavior during depositions, expect more precise testimony about data sources, licensing flows, and internal controls. If questioning is narrowed, it may slow how quickly plaintiffs can surface specifics.
Either way, your catalog is your leverage. The cleaner your records, the stronger your position if your work is scraped, sampled, or embedded into training sets without permission.
Practical steps to protect your catalog
- Register your works. It's the foundation for statutory damages and timely enforcement. If you've delayed, start with your highest-value pieces. U.S. Copyright Office registration
- Lock down licensing terms. Add clear language that restricts AI training use without explicit approval. Include audit rights and takedown cooperation where possible.
- Keep a living inventory. Track publication dates, URLs, editions, co-authors, and licensees. Store original files and hashes to prove ownership and first publication.
- Publish usage policies on your site. State what is allowed for bots, scrapers, and model training. Back it with technical signals (robots.txt, meta tags) while knowing they're not a shield by themselves.
- Monitor model outputs and datasets. Save screenshots, queries, and logs when you find overlaps with your work. Preserve timestamps and sources for potential evidence.
- Update contributor contracts. If you manage a team, align everyone on AI-use permissions, disclosure, and indemnities to avoid gaps.
- Coordinate with peers. Collective action-through guilds, associations, or counsel-adds pressure and reduces duplicated effort.
If you're ever pulled into a deposition
- Prepare with counsel using your actual records: registrations, licenses, and publication history.
- Answer what's asked, stay factual, and don't guess. It's fine to say you don't recall rather than speculate.
- Take your time. Ask for clarifications or breaks through your attorney when needed.
What to watch next
Expect more disputes over how depositions are conducted and how far questions can go into data sourcing and internal practices. The core issue-whether and how AI developers can use protected works for training-remains unsettled and will likely move through multiple courts before a clear standard emerges.
Build your edge while you protect your rights
You can protect your IP and still work smart with AI. If you're exploring tools that serve writers (without giving away the store), start with vetted options and clear use cases.
- AI tools for copywriting - curated options to speed drafts, outlines, and edits while you keep control of the final voice.
Bottom line: tighten your paperwork, set clear permissions, and keep receipts. The legal fights will continue, but your preparation can pay off today.
Your membership also unlocks:
