New AI ownership commercial agreements require human authorship for copyright protection

The lack of copyright protection for purely AI-generated content forces streaming firms to shift from traditional IP law to strict contractual protections. This creates a fragmented landscape where ownership is enforceable between signing parties but potentially vulnerable to third-party infringement. As the industry integrates generative tools for production, the distinction between human-authored expression and AI output will dictate the valuation of digital libraries. The broader ecosystem must now reconcile these U.S. standards with the extraterritorial reach of the EU AI Act, which imposes transparency requirements on any AI output used in the European market. Watch for the outcome of Andersen v. Stability AI to set the precedent for how training data provenance affects downstream licensing liability.

The legal framework governing AI-generated content is tightening across multiple jurisdictions, with direct implications for companies like Anthropic and Stability AI that supply generative models used in media production. In February 2025, the U.S. Copyright Office published Part 1 of its report on digital replicas, concluding that AI-generated outputs lacking sufficient human authorship remain ineligible for copyright registration, reinforcing the human-authorship standard that now shapes contract drafting across the streaming and entertainment sectors. The office followed with Part 2 in January 2026, which addressed copyrightability of AI-assisted works and introduced a spectrum-based analysis for determining when human creative contribution crosses the threshold of protectable expression. That guidance has prompted entertainment studios and streaming platforms to renegotiate vendor agreements with explicit clauses delineating which portions of AI-assisted workflows qualify for protection.

On the litigation front, Stability AI faces a pivotal test that will shape how training-data provenance affects downstream licensing. In June 2025, a federal judge in the Northern District of California denied Stability AI's motion to dismiss in Andersen v. Stability AI, allowing the plaintiffs' direct copyright infringement claims to proceed to discovery, a ruling that legal analysts say could establish precedent for whether model training on copyrighted works constitutes infringement or fair use. Separately, the EU AI Act's transparency obligations, which began phased enforcement in August 2025, require providers of general-purpose AI models to publish training-data summaries and comply with copyright opt-out mechanisms. The European Commission published its first set of guidelines on general-purpose AI model obligations in July 2025, specifying that model providers must demonstrate compliance with the Text and Data Mining reservation under the EU Copyright Directive, a requirement that directly affects how Stability AI and Anthropic license training corpora for European deployments.

Technical and market developments are adding urgency to these contractual shifts. Anthropic released Claude 4 in May 2025 with expanded multimodal capabilities including image generation, prompting the company to update its commercial terms of service to clarify that customers retain ownership of outputs while Anthropic retains a license to use outputs for service improvement. That clause structure has become a template that streaming companies are now adapting for their own vendor agreements. Meanwhile, Stephen Thaler's ongoing efforts to secure patent and copyright recognition for AI-generated works continue to test the boundaries of the human-authorship requirement; the U.S. Court of Appeals for the Federal Circuit affirmed in 2024 that the Patent Act requires a human inventor, rejecting Thaler's argument that an AI system could be listed as sole inventor, a decision that has pushed companies to ensure their contracts explicitly assign human contributors to any AI-assisted creative process.