Artificial Intelligence and IP: Navigating the Legal Challenges

The huge surge in press, public and commercial interest in generative AI tools has generated a parallel surge of interest in IP ownership and exploitation issues. There is a new urgency to questions around IP ownership and infringement: can content or inventions made by AI tools be protected by IP rights and, if so, who owns them? If the source material is protected by IP, are those rights infringed - either by use of the source material to develop, train and run an AI tool, or by its output?

To date much of the focus has been on whether an AI tool can create and own IP in its own right. In most (although not all) jurisdictions, under the current law, the answer is no: an AI tool cannot be the inventor of a patented invention or the author of a copyright work. However, this isn't the only issue, or even the most important one. As a wide range of businesses begin to implement and innovate using AI generative tools, the entire AI ecosystem is now being viewed through an acutely commercial lens. Whether and to what extent the training and output of AI models infringe IP rights are now front and centre of the discussion.

The answer to the AI ownership question will not be the final word, for a number of reasons. As things currently stand, an AI tool has no legal personality and so can’t transfer or enforce any IP rights it does own, or be held liable for infringing the IP rights of others. Many of the tools being described as AI require a level of human input to create content, and AI-assisted human creations can in principle be protected by, or infringe, IP. Finally, absent express provisions to the contrary, existing legal tests for IP subsistence and infringement will still apply. Determining the IP position therefore requires courts and IP offices to apply concepts developed in the world of human intelligence - like creativity, originality and inventiveness - to the developing world of artificial intelligence.

There is also huge scope in this space for legislative, policy and technological developments (not to mention case law in those markets). The winds of change can blow rapidly: for example, the UK took a pro-AI stance in August 2022 with a proposal to adopt a broad exception to copyright for text and data mining (TDM), which would facilitate the use of copyright-protected content in AI training datasets, but by February 2023 had apparently dropped those proposals, citing the difficulty of striking the right balance between the creative, digital and AI sectors.

These questions are likely to keep courts, legislators and lawyers busy for the foreseeable future. The key legal and factual battles are likely to play out against the following high-level issues.