AI & Copyright Law
AI copyright law addresses three intersecting questions: (1) whether AI-generated works qualify for copyright protection; (2) whether training AI on copyrighted works infringes those works; and (3) whether AI-generated outputs that resemble existing works constitute infringement. The U.S. Copyright Office has issued guidance confirming that works generated entirely by AI without human authorship are not copyrightable, while human-AI collaborative works may receive protection for the human-authored elements. Whether training constitutes infringement or qualifies as fair use remains unresolved in appellate courts.
AI copyright questions affect every segment of the entertainment industry simultaneously. Music supervisors clearing songs must assess whether AI-generated music infringes existing works. Film attorneys drafting production agreements must address ownership of AI-generated scripts and visual elements. Rights holders must understand their options when AI companies train on their catalogs without permission. The financial stakes are extraordinary — if training without licenses is found to infringe, AI companies face retroactive liability measured in billions.
Requires human authorship for copyright protection — the foundational rule applied to AI-generated content.
Fair use — the primary defense asserted by AI companies against copyright infringement claims for training data.
Prohibits removal of copyright management information — relevant when AI systems strip metadata from training data.
The Copyright Office confirmed that AI-generated content requires human authorship for registration. Hybrid human-AI works may be registered for their human-authored elements only. Provides case-by-case registration guidance.
Whether AI image generators trained on artists' copyrighted works without license infringed those works.
Direct copyright infringement claims survived early dismissal — the most significant AI training data case currently in litigation. Trial expected 2026.
Whether Stability AI's training on Getty's image database without a license constituted copyright and trademark infringement.
Case is in discovery. Getty's claims survived dismissal and represent the most commercially significant pending AI infringement case in terms of potential damages exposure.
Whether a work generated entirely by an AI system without human authorship qualifies for copyright protection.
The court affirmed the Copyright Office's refusal to register a fully AI-generated work — establishing that human authorship is a constitutional and statutory prerequisite for copyright protection in the United States.
Whether photographs could be copyrighted as works of authorship requiring human creativity.
The Supreme Court established that copyright requires human authorship and intellectual conception — the foundational precedent now applied to determine that AI alone cannot be an author.
The AI copyright landscape is creating a two-tier economy: rights holders with large, high-quality catalogs have leverage to negotiate AI training licenses and establish market value for their content, while individual creators whose works were scraped have limited individual recourse and must rely on class action litigation. Studios face conflicting pressures — as both rights holders (whose content is being used to train AI) and technology adopters (who want to deploy AI tools in production). The uncertainty is driving both licensing deals and litigation simultaneously.
Register copyrights in client catalogs now — registration is a prerequisite to statutory damages in infringement suits, and the window to register before potential AI litigation is narrow.
When a client uses AI tools to generate creative content, document the human creative choices made in the process — these records will be essential if copyright ownership is challenged.
Advise entertainment companies acquiring AI-generated content to obtain representations and warranties about the AI system's training data — indemnification provisions should cover training data infringement claims.
For clients whose content has been used to train AI systems without permission, evaluate whether class action is the appropriate vehicle — individual claims are rarely economically viable given litigation costs.
Monitor the Andersen and Getty cases closely — the first appellate ruling on AI training data will fundamentally reshape advice on licensing, clearance, and IP portfolio strategy.
Copyright protection requires human authorship — AI alone cannot be an author, and fully AI-generated works enter the public domain.
Human-AI collaborative works may be registered for the human-authored elements only — the extent of human creative control is the critical factual question.
Whether AI training on copyrighted works constitutes fair use is the most significant unresolved copyright question in current entertainment law.
Commercial stakes are enormous — if training requires licenses, every major AI company faces potential retroactive infringement liability.
Licensing is emerging as the practical compliance standard as AI companies seek to avoid litigation risk from training data claims.
The human who made sufficient creative choices in the process. If a human author selected, arranged, and edited AI outputs to express their own creative vision, the human-authored elements are protectable. If the human merely entered a prompt and accepted the output, protection is uncertain at best.
Only if a human author exercised sufficient creative control. A script generated by entering a few prompts likely cannot be copyrighted as a work made for hire — it may not be copyrightable at all. One heavily edited and revised by a writer may qualify for protection of the human-authored elements.
At minimum: who owns AI-generated materials, whether AI tools may be used in the creative process, what disclosure obligations apply, and how guild compliance is handled. Sophisticated agreements also address training data warranties, output restriction representations, and indemnification for training data infringement claims.