The best case for a UK opt-out was matching the EU. It was right to drop it anyway

The UK was right to drop its AI copyright opt-out. But the best argument for it was never "AI firms want it". It was "Europe already does it", and that argument deserves a proper answer, because it will come back.

The news hook

On 18 March 2026 the UK government's report on copyright and AI said "a broad copyright exception with opt-out is no longer the government's preferred way forward." Of 11,520 consultation responses, 3% backed the opt-out option. We covered the decision, and how musicians won it, in The UK drops its AI copyright opt-out plan. The report picked no replacement, which leaves the old idea free to return.

The best case for opt-out

Make it the way its supporters did in that same report. The option "would be similar to that provided in EU law by Article 4 of the Digital Single Market Directive", and the report itself says "alignment with EU law was one of its potential advantages." Tech UK said some of its members saw it as the "minimum bar necessary" to stay competitive with the EU and globally. The IP Federation wanted it enacted "without delay" so the UK stays a leading place for AI investment. The Law Society said "greater clarity of the law was needed".

The EU part is not spin. Article 4 of the EU's 2019 copyright directive requires member states to allow text and data mining of "lawfully accessible works", on condition the use "has not been expressly reserved by their rightholders in an appropriate manner, such as machine-readable means" for content online. That is an opt-out. A model trained lawfully in Paris on unreserved music could be unlawful to train in London. For a UK music-tech startup, that gap is a reason to build somewhere else, and an unclear law helps nobody, including artists who want to license.

That's a fair argument. We think it misreads what the EU actually built.

What the EU added that the UK didn't

The EU didn't stop at the opt-out. Its AI Act Article 53 requires providers of general-purpose AI models to put in place a policy to identify and comply with "a reservation of rights expressed pursuant to Article 4(3)" of the copyright directive, and to "make publicly available a sufficiently detailed summary about the content used for training", on a template from the AI Office.

That second duty is what gives an opt-out any teeth. Without it, a songwriter who reserved their rights has no way to find out whether a model ignored the reservation. The UK report, by contrast, only proposes to "work with industry and experts to develop best practice on input transparency" and to watch other countries' rules. So "align with the EU" in the UK debate meant importing Europe's exception without Europe's disclosure law. That is not alignment. It is the half that suits developers.

What we think

Reverse the order. If alignment is the goal, the UK should first legislate the part of the EU model it hasn't got: a statutory duty to publish a training-data summary, with a template detailed enough that a rights holder can check it. That gives AI firms the legal clarity they asked for, gives musicians the one tool every opt-out or licensing scheme depends on, and costs the government nothing it has ruled out.

Only then is there an honest debate about exceptions. Our guess is that once disclosure is law, the case for an opt-out gets weaker, not stronger, because licensing becomes easier to price when everyone can see what was used. That is our reading, not something the report or the EU texts say.

What would change our mind: evidence that the EU's summaries, as actually published, are too vague for a rights holder to act on. If the transparency half doesn't work in practice in Europe, alignment is worth even less, and the UK should hold out for licences in every case, which is what 81% of respondents chose as their preference, according to the government's report.