Are AI art tools allowed to ‘scrape’ the web for images?

On 17 January 2023, Getty Images brought copyright infringement claims in both the US district (Delaware) court and in the UK High Court against Stability AI, the company behind ‘Stable Diffusion’ which is one of the artificial intelligence-powered image generators which have emerged over recent months.

AI generators are able to develop images in response to keywords or instructions entered by a user. The AI generator must be trained over time to associate images with keywords, and this training relies on vast amounts of data which are often pulled from the internet. Getty makes a huge number of photographs and images available on the internet for users to view and purchase. In its claim, Getty alleges that Stable Diffusion unlawfully copied millions of its copyright-protected images and the associated metadata from their databases. In the UK case, the main legal issues which are expected to arise as a result of this case involve questions of fair dealing exceptions under UK copyright legislation (Sections 29 and 30 of the Copyright, Designs and Patents Act 1988). 

However, this specific type of use has not yet been considered by the UK courts. The government has already called for consultations on the topic and there have been proposals to expand the current ‘text and data mining’ exception beyond ‘non-commercial research purposes’ to cover all purposes. The outcome of this would allow the scraping, trawling and mining of data (including images) found on the internet to train AI systems free of charge, without seeking permission from the rights holders. As expected, this announcement led to a lot of public scrutiny and is now being re-considered pending an impact assessment for businesses in the creative industries. The consultation and its outcome is likely to have an impact on the eventual outcome of the Getty case.

Comment

We find ourselves once again in the unchartered waters of internet copyright law. As with the early days of digital music, where Napster for example lost its legal battle against five of the world’s largest record labels, the question on the extent of unauthorised use of copyright artwork will be considered by the courts.

This case is likely to set an important precedent and will assist with the developing legislative framework in this area of law; providing clarity to artists, media companies and end-users who may have wanted to rely on and use the artwork generated by the AI powered tool. A balance must be struck between protecting artists and media companies, whilst allowing the development and training of AI systems at little or no cost.