Google Declares AI Training Is Fair Use — Putting It on a Collision Course With the Music Industry
Google's new policy paper argues AI training is fair use and copyright should govern outputs, not inputs — a direct threat to the industry's legal strategy.
· 4 min read
The Stakes
If Google gets its way with regulators, the music industry's entire legal and lobbying campaign to extract licensing revenue from AI companies collapses at the foundation. Google has now put that position in writing.
What Happened
On June 25, 2026, Google published a 21-page policy paper titled A Pragmatic Approach to AI Governance in America, authored by Kent Walker, the company's President of Global Affairs. The document explicitly argues that training AI models on publicly available web data is a "transformative, non-expressive use" that should remain protected under fair use in the US, and under text-and-data-mining exceptions abroad.
The paper uses a telling analogy: Google compares AI training to "an art student taking inspiration from walking through a gallery." Music Ally's Stuart Dredge noted the analogy has an obvious flaw: galleries involve payment and permission, revenue from sales is shared with artists, and works are bequeathed willingly — the gallery didn't "wander into their house and take them."
On where copyright enforcement should actually apply, Google is unambiguous. The paper says copyright concerns raised by generative AI are best addressed at the level of outputs, not inputs — whether a specific piece of content copies an existing work, rather than how the model was trained. On enforcement, Google argues that technical filters should not "automate subjective decisions like whether something is 'too similar' to a prior work," and that infringing material is best handled through standard notice-and-takedown systems.
This is not a new position for the company. In 2023, Google made the same argument in a filing with the US Copyright Office, calling AI training a transformative fair use and saying courts — not new legislation — should resolve the question.
Google does offer some concessions in the paper. It says responsible developers should give website owners control over whether their content is used for model development through machine-readable tags such as its own Google-Extended control, and that it is "exploring new types of partnership and value-exchange models" with rights holders, including paying for access to specialized non-public content. Google also points to its support for the NO FAKES ACT as evidence of good faith on creator protections.
Why It Matters
The music industry is fighting this exact legal question on multiple fronts simultaneously, and Google's paper is a direct shot across the bow.
The RIAA, on behalf of Universal Music Group, Sony Music, and Warner Music Group, sued AI music platforms Suno and Udio for "mass infringement" of copyright in mid-2024. Music publishers have separately sued AI companies including Anthropic. Independent artists have also sued Google directly, claiming it used YouTube's own catalog to train its Lyria 3 AI music tool.
The industry's legal theory depends on winning the input argument — establishing that ingesting copyrighted recordings and compositions without a license is itself an infringing act. If regulators accept Google's framing and shift the legal lens entirely to outputs, that theory dies. Labels and publishers would be left chasing individual infringing tracks through notice-and-takedown — a whack-a-mole process — rather than collecting upstream licensing fees from model developers.
The scale problem makes Google's takedown proposal look particularly thin: according to Deezer's latest data, 75,000 fully AI-generated tracks are now being uploaded to streaming services every single day, up from 10,000 a day at the start of 2025. Standard reporting and takedown mechanisms were not built for that volume.
There is also a market-structure angle worth noting. Udio, one of the companies the RIAA sued, has since moved away from defending its training under fair use and instead signed licensing deals with Universal, Warner, Merlin, and Kobalt — suggesting that at least some AI companies see licensing as the more viable long-term path. Google, with far greater resources and political reach, is betting it can define the rules before that market norm solidifies.
The Takeaway
Google's paper is lobbying dressed as policy analysis, and the music industry should treat it as such. The company has now staked out its position publicly and in detail, giving policymakers a concrete framework to adopt or reject. The industry's response needs to be equally concrete — not just legal filings in individual cases, but a coherent regulatory argument that addresses the output-versus-input question head-on and accounts for the volume of AI-generated content that notice-and-takedown was never designed to handle. The next arena is Congress and the Copyright Office, and Google just fired the opening shot.
Related: A Gambler Rigged Spotify's Charts to Win a Bet, and the Artist Paid For It
Related: Bob Vylan Are Suing the BBC. This Is How an Artist Fights Back Against a Broadcaster's Framing.
Related: The Music Industry Just Agreed on How to Label AI. The Streaming Services Did Not.
Related: Suno's Leaked Source Code Names the Streaming Services It Scraped
Related: States Demand Discovery Into White House Talks Behind the Live Nation Settlement
Sources
The Newsletter
The music business, decoded.
Join industry professionals who read our briefing on deals, royalties, and the business behind the music. No noise, no spam.