Legal Considerations
Session 13.4 · ~5 min read
This Is Not Legal Advice
This session maps the current legal terrain for AI-generated content. It is not a substitute for consulting an actual lawyer, especially if you are producing commercially valuable content at scale. The law is evolving, varies by jurisdiction, and has significant unresolved questions. What follows is what we know as of early 2026.
The Core Question: Who owns AI-generated content? The short answer: purely AI-generated content with no meaningful human creative input is likely not copyrightable in the US. Content substantially directed by a human, with AI as a tool, likely is. The boundary between these two categories is being defined through ongoing litigation.
Copyright and AI Output
In March 2026, the U.S. Supreme Court denied certiorari in the Thaler case, reaffirming that works without a human creator are ineligible for copyright protection under current U.S. law. This settled one end of the spectrum: purely AI-generated works, with no human creative involvement, cannot be copyrighted.
The U.S. Copyright Office released Part 2 of their AI report addressing copyrightability in 2025. The key finding: human-driven selection, arrangement, and coordination in the creative process can constitute sufficient human authorship to merit copyright protection, even when AI tools were used in production.
No human creative input
'Generate an article about X'"] --> B["Gray Zone
Human direction + AI execution
'Generate using these specs,
then human edits significantly'"] B --> C["Human-Authored with AI Tools
Substantial human creative control
'Human structures, AI drafts,
human revises voice/content'"] A2["NOT copyrightable
(US, as of 2026)"] -.-> A B2["Likely copyrightable
(depends on level
of human involvement)"] -.-> B C2["Copyrightable
(AI as tool,
like Photoshop)"] -.-> C style A fill:#c47a5a,color:#111 style B fill:#c8a882,color:#111 style C fill:#6b8f71,color:#111 style A2 fill:#c47a5a,color:#111 style B2 fill:#c8a882,color:#111 style C2 fill:#6b8f71,color:#111
The practical implication for your pipeline: the more human creative direction your process involves (choosing the angle, structuring the argument, editing for voice, adding original insights), the stronger your copyright claim. A pipeline that prompts AI for a first draft, then substantially revises it, is in a stronger legal position than a pipeline that publishes raw AI output.
Terms of Service
Every AI API has terms of service that define who owns the output. Read them. The major providers as of 2026:
| Provider | Output Ownership | Commercial Use | Key Restriction |
|---|---|---|---|
| Anthropic (Claude) | You own the output | Permitted | Cannot use output to train competing models |
| OpenAI (GPT) | You own the output | Permitted | Cannot use output to train competing models |
| Google (Gemini) | You own the output | Permitted under API terms | Terms vary between consumer and API products |
These terms can change. Review them quarterly. The ownership grant in the ToS is a contractual right, which is separate from copyright. You can own the output contractually (the provider will not claim it) while the output may or may not be copyrightable under the law. These are two different questions.
Fair Use and Training Data
The Copyright Office concluded in May 2025 that some uses of copyrighted works for AI training qualify as fair use, and some do not. Specifically, AI developers who use copyrighted works to train models that generate "expressive content that competes with" original works are going beyond fair use.
This matters for you if your content is used as AI training data (you cannot prevent this with current technology, though robots.txt and opt-out mechanisms exist) or if you are concerned about the provenance of the model's training data.
Practical Steps
- Read the ToS for every AI service you use. Document ownership terms, commercial use permissions, and restrictions.
- Maintain evidence of human creative involvement. Your production logs, prompt iterations, editorial changes, and review notes all demonstrate human direction.
- Register copyright for commercially important works. If a piece of content is valuable enough to protect, register it with the Copyright Office. The registration process requires you to identify the human-authored elements.
- Do not publish raw AI output as your own without significant human editorial involvement. Beyond the ethical considerations from Session 13.3, this protects your legal position.
- Consult a lawyer for high-value projects. If you are publishing a book, launching a course, or producing content for a large client, get legal counsel familiar with AI and intellectual property.
Jurisdiction Matters
Copyright law varies by country. The EU AI Act has different disclosure and transparency requirements than US law. If your content reaches international audiences or you publish through platforms governed by different jurisdictions, the legal picture is more complex than any single-country analysis suggests. This is another reason to consult local legal counsel for commercially significant content.
Further Reading
- Copyright and Artificial Intelligence, U.S. Copyright Office
- Copyright and AI Part 2: Copyrightability Report, U.S. Copyright Office (2025)
- An Update on AI Copyright Cases in 2026, Norton Rose Fulbright
- AI Copyright Lawsuit Developments in 2025: A Year in Review, Copyright Alliance
Assignment
Read the terms of service for every AI API you use. For each, document: who owns the output, what you can and cannot do with it, and any restrictions on commercial use. Then evaluate your current pipeline against the copyright spectrum from this session: where does your process fall? How much human creative involvement exists at each stage? Keep this as a reference document and flag anything that concerns you for discussion with a legal professional.