What is the potential impact of Getty’s lawsuit against Stability AI on the music industry?
MBW Views is a series of exclusive commentaries/reports from music industry luminaries… with something to say.
In the following article/editor, Jonathan Coote (pictured), associate and music lawyer at Bray & Krais Solicitors, examines tThe potential impact of the Getty lawsuit against Stability AI on the music industry and suggests that there needs to be “urgent clarification of the law relating to AI training in the UK”.
From the allegations that the AI music platform Suno is capable of being trained on copyrighted recordings ARRIVE artists call on tech platforms to stop infringement and dumping rights, there remains deep concern in the industry that copyright-protected music is being used without license or appropriate compensation to train AI tools that can copy (and eventually replace) so) music is created by humans.
But what is the legal framework to prevent such training?
In the UK, current law appears to prohibit training AI tools on copyrighted works for commercial purposes.
The government initially refused to enact the EU's softer “opt out” approach during the Brexit transition period, but then sought to fully transform the law in 2022 by attempting to introduce a extremely permissive exceptions.
This plan was quickly abandoned after being described by the copyright holder as “music laundering”. The government's next initiative is to establish a working group to develop a system by which AI companies can obtain a “fair license.”
Perhaps unsurprisingly, given the diametrically opposed interests of tech companies and copyright holders, these negotiations failed.
Unlike in the US, where “fair use” protections can consider the broader impact on the public, UK copyright protections are narrow and technical. .
This means that applying existing law (especially that created before large language models and general AI) may not be enough to address the economic, innovative and social consequences of AI . Without further legislation, this could create unhelpful precedents.
While the limitations on training under UK law are clear on the face of it, the outcomes of practical cases are not, as they can be very fact specific. There is an important case that could shape UK law on these issues: stock image site Getty Images sued Stability AI in the US and UK for training its models on databases. Getty image data. In the UK courts, Stability AI failed to settle the case but its defense highlights some of the difficulties in bringing a successful infringement claim.
First, arguments that an individual work is infringing are inherently difficult to prove because the work must reproduce “substantial parts” of a particular work, and AI tools are often designed with Safeguards prevent this. Furthermore, Getty must prove that Stability AI committed the infringement in the UK. In its defense, Stability AI denied this, claiming that the developers involved were overseas and that any alleged copying occurred on foreign cloud servers. Getty also argued that the model in question, Stable Diffusion, was an imported infringing “article,” typically used for physical goods such as pirated CDs. The effectiveness of these arguments remains to be seen.
There are also some very specific elements about Getty's case that many in the music industry cannot rely on. First, Getty attempted to reproduce distorted watermarks in the output images, proving that its images were clearly used during training (a difficult task due to its “boxy” nature). black” of most AI tools) and thus lays the groundwork for trademark infringement. Second, because Getty has compiled a library of stock images, it can also claim violation of its “database rights,” which is unlikely to be effective against most copyright owners. permission. Because of the technical and legal complexity, if the case ultimately leads to a decision it will be difficult to establish a universal and definitive precedent.
With AI-generated music already being distributed on streaming services, music companies and musicians are understandably worried about its popularity. However, while the music industry has criticized AI companies for “music laundering,” with one notable exceptionit was reluctant to bring proceedings.
Furthermore, the music industry is perhaps more worried than other industries about the impact of deepfakes, given recent astonishing developments in voice cloning technology. Although many people instinctively feel that this is illegal, technically there are no specific image or likeness rights (although they are widely licensed) that would explicitly limit clarify the outputs of such tools in the UK. Therefore, the most effective way to sue may be to claim that the training itself was infringing. Therefore, training is extremely important for the industry.
The regulatory landscape is rapidly evolving. The US likely to introduce legislation to create new “digital representation” rights to tackle deepfake problem and, across the channel, The EU AI Act is set to increase transparency on requests from AI companies operating in the EU for their training data, and could also extend “opt-out” restrictions beyond territory.
There are a number of positive initiatives underway in the UK: one The all-party parliamentary group is investigating the impact of AI on the music industry (you can find the author's thoughts on this This) And A bill in the House is bringing real regulation to Congressincludes important transparency provisions.
However, given the potential inadequacies of the current law, the UK government needs to urgently take advantage of this work to provide much-needed clarity.
Music business worldwide
