The "Wild West" era of scraping British music to train algorithms is officially over. On December 16, 2025, the UK government released its Statement of Progress on Copyright and AI, delivering a decisive regulatory victory to rights holders that effectively kills the argument for open-access model training in the region.
For two years, the tech sector has lobbied for a "Text and Data Mining" (TDM) exception—a legal loophole that would have allowed AI developers to ingest copyrighted music without permission or payment. The government has now firmly rejected that path, signaling that the future of British AI will be built on licensed, transparent datasets rather than scraped archives.
The statistical landslide
The government’s reversal wasn't subtle; it was driven by overwhelming data. The consultation results painted a picture of a unified creative sector and a tech industry unable to make a compelling public interest case.
- 95% of respondents supported a licensing-based environment.
- Only 3% backed the government's initial proposal for a broad TDM exception.
This statistical rout made the initial 2022 proposal—which sought to allow mining for any purpose—politically toxic. The government has conceded that "extracting informational value" from songs constitutes a copyright-relevant act, validating the business models of major rights holders like Universal and Sony.
A pricing reality check
This shift has immediate balance sheet implications for the burgeoning AI music sector. Startups operating in London can no longer rely on the "fair use" defense that is currently being litigated in US courts.
Consider Mirelo, which recently raised $41M in seed funding. Under a TDM exception, that capital could have been poured almost exclusively into engineering and user acquisition. Now, a significant tranche must be allocated to legal data acquisition. The "move fast and break things" model has hit a regulatory wall; the new model is "move carefully and pay for things."
Industry vindication
UK Music CEO Tom Kiehl didn’t mince words, calling the findings a "vindication" of the sector's aggressive lobbying efforts. With the UK music industry valued at roughly £8 billion, the government realized it was risking a proven economic engine to subsidize a speculative one.
Key insight: The new framework moves beyond just permission; it prioritizes transparency. Rights holders will likely gain the ability to audit AI models to see if their catalogs were ingested during the training phase.
Global regulatory drift
This decision clarifies the UK's position on the global stage. While the US legal system is bogged down in high-stakes litigation (like UMG v. Anthropic) to define the boundaries of fair use, the UK is legislating clarity.
By rejecting TDM exceptions, the UK aligns itself closer to the European Union's AI Act—which mandates detailed summaries of training data—and away from the American "fair use" ambiguity. For global labels, this creates a transatlantic leverage point: they can now use the UK’s strict compliance standards as a baseline for global licensing deals.
The strategist's playbook
With the legal threat of non-consensual scraping diminished in the UK, rights holders should pivot from defense to offense.
- Audit your paper: Ensure management and recording contracts explicitly define "AI training" as a licensable right separate from streaming or sync.
- Prepare for transparency: Expect new tools mandated by Section 137 of the Data Act that will require AI firms to disclose their inputs. Labels need the technical infrastructure to verify these disclosures.
- Price the data: The government has confirmed music is data. The next challenge is establishing a rate card for training licenses that captures value without stifling the tools that artists actually want to use.