The IP in AI – Recent Updates and Developments

While Australia does not currently have any AI-specific regulation, it has demonstrated its eagerness to reform in this area during the early months of 2024. On 17 January 2024, the Australian Government released its interim response to the public consultation for its ‘Safe and Responsible AI in Australia’ Report.⁵ In response to its findings, the Minster for Industry and Science also announced a new Artificial Intelligence Expert Group created to advise on ensuring “transparency, testing and accountability” of AI systems.⁶

In parallel, in December 2023, the Australian Government established a Copyright and Artificial Intelligence Reference Group to “better prepare for future copyright challenges emerging from AI”.⁷ The initial focus of the Reference Group is the use of copyright material as inputs into AI systems, while future topics may include both copyright infringement by, and copyright protection of, outputs from AI systems.

Though the outcomes of such reports and committees remain purely theoretical at present, the Australian Government has demonstrated its commitment to policy reform in its recent 2024-2025 budget – which announces it will provide AUD$39.9 million over five years into developing policies to support the safe use of AI technology.⁸

During 2024, litigation concerning both AI inventorship and infringement of IP rights through AI systems has continued in the UK Courts. In particular, the final decision in the UK Thaler test case was recently determined by the UK Supreme Court – see here for summaries of the previous decisions. In the Supreme Court case,⁹ it was held that only a natural person can be recognised as a patent inventor and thus the AI system (DABUS) could not be entitled to inventorship rights.

The UK High Court is also set to hear argument in Getty Images v Stability AI – a case concerning the Stable Diffusion text-to-image tool alleged to have used ~12 million of Getty Images’ photos for its training purposes without permission – during summer 2025.¹⁰ This case was brought in 2023 (see previous summary here), with Stability AI filing its first defence in March 2024. The defence alleges not only that the potential ‘copying’ of the works took place outside the UK, but also that any images created by the AI based on text prompts does not recreate “any part of the copyrighted works” and that any copying that results from the insertion of an unlicensed image prompt by the user is not attributable to Stability AI. Alternatively, it argues it may rely on the pastiche exception found in the Copyright, Designs and Patents Act 1988.  

The E&W High Court in Emotional Perception¹¹ also recognised artificial neural networks (ANNs) as patentable subject matter (see here for our previous blog post). This case concerned a patent application for an ANN, which ingested music tracks and used information including human perceptions and descriptions of the music to recommend similar tracks. While the Comptroller of Patents had held that such systems fell within the s.1(2)(c) exclusion of computer programs from patentability, Sir Anthony Mann found that the ANN software was not a computer program as it was not implementing human-controlled code by “operating at a different level” than the underlying software.¹² This decision provides some support for the patentability of AI software, which is of particular relevance for companies involved in creating or investing in artificial intelligence tools.

In a similar vein to the UK litigation concerning AI infringement of copyright, the US has also seen a rise in creatives (predominantly writers and actors) commencing lawsuits against generative AI companies such as Open AI, ChatGPT and Microsoft. The most recent filing in this long list of complaints was made on 30 April 2024 by eight US newspapers (New York Daily News, Chicago Tribune, Orlando Sentinel, Sun-Sentinel, Mercury News, Denver Post, Orange Country Register and Pioneer Press) against Microsoft and Open AI.¹³ The plaintiffs allege that the Defendant “purloin[ed] millions of the Publishers’ copyrighted articles without permission and without payment to fuel the commercialization of their generative artificial intelligence products”.¹⁴ While most of these cases remain at the discovery stages, such decisions will be crucial to determining how AI may be seen to infringe copyright in the future. 

This is also particularly important in light of recent developments in deepfake AI technology, with Scarlett Johansson recently alleging that OpenAI copied her voice for its startup and other voice actors also threatening class actions against Lovo Inc for misappropriating their voiceovers using AI.¹⁵ While these issues remain un-litigated at present, they would present a significant expansion of the copyright protection afforded to individuals in light of AI technologies.

In terms of copyright protection afforded to AI-generated outputs, some recent decisions of the Copyright Office have provided early insight into the US position on the issue. In both Zarya of the Dawn and Theatre D'opera spatial, the US Copyright Office refused to recognise either AI-generated work as the product of human authorship, denying them both copyright registration. In Zarya, the Office held that the since the user (Ms. Kashtanova) relied heavily on the unpredictability of the images produced by the AI software Midjourney, such that her prompts did not evidence sufficient control or guidance over the outputs, they did not meet the requirement for human authorship.¹⁶ More recently, the Copyright Office also denied registration to Theatre D'opera spatial – an award-winning image generated by the artist Jason Allen using Midjourney. Despite Allen arguing that the image generation involved the use of at least 624 prompts and that he edited the output subsequently in Photoshop, the Copyright Office held the work contained more than a de minimus amount of AI content. Since Allen also refused to disclaim the AI contributions, the Office denied the work registration both initially and in a subsequent request for reconsideration.¹⁷

Beyond these initial decisions, the US approach is also expected to be clarified over the coming months in the 2024 US Copyright Office Report focusing on the impact of generative AI on policy. US Congress has also recently, as of April 2024, attempted to address such issues by introducing legislation which would mandate the disclosure of the copyrighted materials AI companies use to develop generative models. The Generative AI Copyright Disclosure Act would require companies to submit documents containing the copyrighted works used in training datasets to the Copyright Office at least 30 days prior to the release of their AI software. This would enable the Office to create a public database of the notices, such that creators could search and determine if their work has been included without a licence.

The first ruling in the European Union on the copyright protection afforded to AI-generated works was handed down in the Czech Municipal Court of Prague during April 2024 (see here for our previous blog post on the decision). The case concerned a claimant who created an image using DALL-E, published it on his website, and began an action for copyright infringement when the defendant used the same image on its own website. The Court ultimately dismissed the case on procedural grounds and, unsurprisingly, held that AI could not be an author of copyright. However, it also left open the possibility that a sophisticated prompt could constitute a “unique creative contribution” which may be sufficient to allow the (human) user to obtain copyright in the AI output. Whilst this decision is far from conclusive on the protection of AI-generated works in the EU, it provides some insight into the importance of user prompts in establishing human authorship of those works.

After first being proposed by the European Commission in April 2021, the text of the EU AI Act was also finally approved by the Council of the European Union (the final step in the legislative process for the creation of EU laws) on 21 May 2024. The culmination of three years of adjustment and debate, the Act is the “first of its kind in the world”¹⁸ – in that it legislates for safe and transparent AI practices alongside the preservation of important rights for individuals. The Act generally approaches AI regulation from a risk-based perspective, whereby high-risk AI practices are subject to stricter regulation. The Act is expected to enter into force twenty days after its publication in the EU Official Journal in the coming days.