The UK government has abandoned plans to introduce a copyright exception for artificial intelligence training, marking a significant policy reversal that favours creative industries over technology companies seeking unfettered access to copyrighted material.
The decision, confirmed this week, scraps the text and data mining (TDM) exemption that would have allowed AI developers to use copyrighted works without permission or payment. The move follows sustained opposition from publishers, authors, musicians, and other rights holders who argued such an exception would undermine their economic interests and creative control.
The government had consulted on the TDM exception in 2022 as part of broader efforts to position the UK as an AI-friendly jurisdiction following Brexit. Proponents argued that regulatory barriers would handicap British AI companies competing against American rivals operating under more permissive fair use doctrines.
However, creative industry groups mounted an effective campaign against the proposal. The Publishers Association, Authors’ Licensing and Collecting Society, and music industry bodies argued that allowing free use of copyrighted material for AI training would constitute a substantial transfer of value from creators to technology companies without compensation.
Under the revised approach, AI developers must negotiate licensing agreements directly with rights holders or their representatives. This aligns the UK more closely with the European Union’s approach under its AI Act, which preserves copyright protections whilst requiring transparency about training data sources.
Business Impact
The decision creates clear winners and losers across the AI ecosystem. Publishers, music labels, and collecting societies gain strengthened negotiating positions and new revenue streams from licensing deals. Several major publishers have already signed agreements with OpenAI and other AI companies, with reported values reaching tens of millions of pounds annually.
Conversely, AI startups and smaller technology companies face increased costs and legal complexity. Licensing negotiations require legal resources and may prove prohibitively expensive for early-stage ventures. Large technology companies with substantial legal departments and financial reserves are better positioned to navigate this landscape, potentially consolidating market power.
The policy also affects the UK’s competitive positioning. Whilst the government maintains that voluntary licensing provides sufficient flexibility, some technology industry representatives warn that regulatory uncertainty could drive AI development to jurisdictions with clearer exemptions. The UK’s approach now sits between the EU’s strict requirements and the US’s broader fair use provisions.
Implementation Questions
Significant practical questions remain about enforcement and scope. The government has not specified how it will monitor compliance or what penalties will apply for unauthorised use of copyrighted material. Rights holders must still demonstrate that their works were used in training, which requires technical capabilities that many lack.
The decision also leaves unresolved questions about works already used to train existing AI models. Retroactive licensing requirements could expose companies to substantial liability, though the government has not clarified its position on historical training data.
Industry observers expect a wave of licensing deals and potentially litigation as rights holders and AI companies test the boundaries of the new framework. The Copyright Tribunal may face increased caseload as disputes arise over licensing terms and fair compensation.
What’s Next
The government has indicated it will monitor how voluntary licensing develops before considering further regulatory intervention. Industry groups on both sides will watch closely whether sufficient licensing agreements emerge to satisfy AI companies’ data needs whilst providing adequate compensation to rights holders.
Parallel developments in the EU and potential US legislation on AI and copyright will influence whether the UK’s approach proves sustainable or requires adjustment. The outcome will shape not only the economics of AI development but also fundamental questions about how intellectual property rights adapt to machine learning technologies.
This policy shift represents a clear prioritisation of creator rights over technological permissiveness, with implications extending far beyond the UK’s borders as other jurisdictions grapple with the same tensions between innovation and intellectual property protection.
