Report from the Financial Times
In Brief – The UK government is reconsidering a proposal that would allow AI developers, including giants like OpenAI and Google, to train AI models on copyrighted materials without obtaining prior consent from rightsholders. After a two-month public consultation generated little support, officials are expected to delay legislation and revisit their approach, meaning an AI bill addressing the issue of AI training is unlikely to appear in the upcoming King’s Speech. The debate has exposed a divide on the issue between the Labor government, which proposed a “data mining” exception that required copyright owners to opt out if they did not want their work used for AI training, and members of the House of Lords Communications and Digital Committee, who call for a “licensing-first” system requiring transparency about training data, positioning the House of Lords as a defender of creators’ livelihoods and winning backing from hugely prominent British artists such as Elton John and Paul McCartney.
Context – The application of copyright law to AI model training is a foundational legal and regulatory issue facing the industry. Generative AI neural networks have been trained on most of the largest content sites on the internet. While the debate over AI training and copyright is global, US courts are likely to resolve the key issue of whether AI training is “fair use” under US copyright law. Two battling district court opinions released last summer illustrate the complexities. The Trump Administration, aggressively promoting AI development, is not likely to restrict training and Congress won’t act. In the EU, with its landmark AI Act, regulators and expert groups play central roles. The copyright section of their General-Purpose AI Code of Practice has proven contentious for a lack of clarity. US friendliness to AI development is pushing other governments to appear investment friendly as well. Like the UK, Australia proposed to make its copyright law more conducive to training, which, like in the UK, generated pushback from creative industries. Japan already has a copyright exception for AI training. Its creative industries are now energized and arguing it should be tightened.