Silicon Valley can pay to train AI on copyrighted work - it just doesn't want to
AI models run on creative labor - journalism, photography, music, design, code. There's more than enough money in this industry to pay for that input. The choice to scrape first and argue later isn't a necessity. It's convenience.
This doesn't have to be a fight. It can be a licensing market.
Copyright is the business model, not a nuisance
Copyright gives writers, visual artists, musicians, and filmmakers ownership of their work. It's how publications charge for access, how labels invest in talent, and how freelancers get paid fairly.
Right now there are more than 107 active copyright infringement cases against AI platforms worldwide. Eighty-seven are in the United States. If you're told US "fair use" makes all of this simple, don't buy it - it's a litigation war zone, not a free pass. If you want a quick primer on fair use, start here: Stanford's Fair Use overview.
The false choice: innovate or pay
We've seen this movie. Radio paid songwriters. Streaming pays rights holders. Stock libraries license catalogs at scale. Search and social have signed news deals in multiple markets.
Training datasets can be licensed too - with reporting, audits, and revenue shares. The barrier isn't feasibility. It's willingness.
What fair compensation can look like
- Collective licensing: Societies represent creators, set terms, and distribute royalties - efficient and standard in music and visual arts.
- Direct deals: Publishers, labels, stock libraries, and studios license catalogs with usage reporting and minimum guarantees.
- Opt-in/opt-out controls: Clear consent signals, respected by platforms, logged in audit trails.
- Provenance tech: Use content credentials (C2PA) to attach tamper-evident metadata that travels with your work. Learn more at the C2PA initiative.
What creatives can do right now
- Register your work with your national copyright office where possible. It strengthens your position in disputes and deals.
- Add "no-train/no-AI" signals where available (robots.txt, platform settings, or metadata). Not perfect, but it sets intent and builds evidence.
- Use provenance tools (watermarks, content credentials) and keep source files and drafts. Paper trails matter.
- Join a collective or licensing body that is negotiating with AI companies. Collective power works.
- Update your contracts with explicit AI clauses: training consent, model-use restrictions, royalties, and reporting.
- Monitor for misuse with dataset search tools and reverse image/text search. File takedowns when necessary.
- Stay informed and upskill so you can leverage AI ethically while protecting your rights. Start here: AI for Creatives.
What AI companies should do (if they're serious)
- Budget for rights as a core training cost, not an afterthought. Pay minimum guarantees and usage-based royalties.
- Respect consent - no scraping from paywalled sources or places that signal "do not crawl."
- Offer clean pipelines: licensed datasets, audit logs, content provenance, and enterprise-grade indemnities.
- Share value via APIs and rev-share programs so creators benefit when their work improves model outputs.
The upside of paying
Licensing buys legal certainty, better data quality, and brand trust. It opens doors with enterprise customers who need clean rights. It also aligns incentives: creators keep making great work, and AI companies train on it without a courtroom shadow over every release.
The bottom line
There's enough money in AI to pay the people who supply its raw material. Creators deserve a slice of the largest wealth creation in tech history. We can build a functioning market - or keep funding a global legal brawl. The smarter move is obvious.
Your membership also unlocks:
