UK AI copyright plans align – and diverge – with the EU position

Frankfurt-based Alexander Bibi of Pinsent Masons highlighted the divergent approaches following publication by the UK government of its statutory reports on AI and copyright last week – reports that Pinsent Masons experts said signal a wider pivoting of UK AI policy.

The UK government confirmed that it no longer views “a broad copyright exception with opt-out” as its “preferred way forward”.

That option, which it expressed a preference for when originally consulting on possible reforms, would have entailed a widening of the existing text and data mining (TDM) exception in UK copyright law for AI training purposes. At the moment, the TDM exception can only be relied upon for non-commercial research. Also envisaged under the option was the introduction of a mechanism to enable rightsholders to opt their content out from being used in AI training and the further introduction of measures requiring AI developers to be transparent about the works they train their models on, so rightsholders had sight of the activity and could decide whether to reserve their rights.

The government confirmed that it has no immediate plans to pursue any copyright reform in relation to AI training. Its approach instead is to undertake more evidence gathering and otherwise maintain a mostly monitoring brief. Bibi said EU policymakers have taken a different path.

“EU law provides a broadly applicable text and data mining exception that enables AI training unless the rights holder explicitly exercises an opt‑out in a machine‑readable format,” Bibi said. “Only then does licensing become necessary for the relevant content. It remains to be seen how major rights holders, such as publishers and collecting societies, will approach this and develop corresponding licensing models. In practice, most large rights holders in the EU already offer licensing schemes tailored to AI training.”

EU copyright law applicable to AI training is set out in three different frameworks. While baseline rights and protections for creatives are contained in the 2001 information society directive, the TDM exception and opt-out requirements are provided for in the 2019 digital single market directive. The 2024 AI Act outlines further transparency obligations specific to providers of ‘general purpose AI’ (GPAI) models.

Earlier this month, the European Parliament passed a resolution (14-page / 164KB PDF) calling on the European Commission to rethink how the EU copyright framework applies in the context of AI.

Among other things, MEPs called for rightsholders to be given “full control” over the use of their content for AI training; for a new mechanism enabling rightsholder opt-outs via the EU’s Intellectual Property Office (EUIPO) to be explored; for transparency requirements to be bolstered; for voluntary collective licensing agreements to be facilitated; and for options for “immediate, fair and proportionate remuneration for past uses of copyright-protected works” in AI training to be examined.

Bibi said: “While the resolution itself has no legal weight, it is relevant in the context of the first review of 2019 directive, which could be a prelude to meaningful changes and welcome clarifications, given the directive was adopted before the era of generative AI. It provides a clear picture of the European Parliament's position on this topic. However, it is likely that there will be significant pushback from industry and member states to the position MEPs have set out, especially in the current context and weariness about the need for competitiveness in the AI world.”

While UK and EU copyright law relating to the use of works in AI training differs, Bibi highlighted how UK policy concerning copyright protections for outputs is now expected to become more aligned in light of comments the UK government made in its recent reports.

Specifically, while the UK government said it will “continue to monitor the use and impact of” protections provided for computer-generated works (CGWs) under the 1988 Copyright, Designs and Patents Act (CDPA), it is proposing to remove those protections “in the absence of evidence of its ongoing value”. A ‘computer-generated work’ is defined in the CDPA as work that is “generated by a computer in circumstances such that there is no human author of the work”.

The removal of copyright protections for CGWs would bring UK law into line with the position in the EU, as well as the US.

Bibi said: “Moving forward, only works created with human involvement will be eligible for UK protection. This approach aligns with the European droit d’auteur principle, which centres copyright protection on human creativity.”

“Practically, this shift raises new questions: how much human input is required to meet the minimum threshold of creativity, and how can such contributions be demonstrated? These issues are likely to become central in future disputes and questions of evidence,” he added.