The rapid advancement of AI continues to raise complex questions about the applicability of intellectual property (IP) laws to AI and AI-generated works. The unprecedented pace of development of this technology is pushing enterprises towards self-governance frameworks founded on ethical considerations. IP remains one of the leading and most contentious issues in respect of AI governance.

In 2025 and beyond, we expect to see governments across the world grappling with balancing strategies aimed at encouraging the development of AI and innovation while, at the same time, attempting to modernize IP and AI legal frameworks to account for AI.

Rights in input data used to train AI models and infringement of IP

A highly debated topic is whether use of copyright-protected materials to train AI models should be considered as an infringement of the underlying copyright. Or should AI models be entitled to create new, derived content "informed" by the training data (as a real person may be having consumed the same source information)? This continues to provide a challenge to legislators worldwide.

In 2025, we expect to see increased regulatory scrutiny of organizations that create or use AI technologies which have been trained using information/data protected by IP rights. Regulators worldwide are now paying greater attention to balancing the benefits of AI against concerns about the protection of IP. By way of example, the Labour government elected in the UK in 2024 pledged to bring forward legislation tackling AI in 2025 and opened a consultation on the issue in December 2024. In the consultation, the government is seeking views on an extension of the express exception for text and data mining (TDM) to allow data mining for commercial purposes, coupled with the ability for rights holders to opt out, which would bring the UK more in line with the EU. The consultation runs until 25 February 2025.

The issue of IP infringement has taken center stage in global legislative discourse where AI models are trained on IP-protected data. While in some cases the right to scrape has been set out contractually between AI models and end-users, many large enterprises have been sued in various countries, in respect of unauthorized scraping of copyrighted work resulting in nuanced questions around fair use through democratized data mining of works in the public domain pending decision before the judiciary. Interestingly, many governments are leveraging AI to detect infringement and mitigate the risks. In a judgment of the Hamburg Regional Court, it was held that even a machine-understandable (vs machine-readable) disclaimer of a website specifically precluding the scraping thereof for the purpose of data mining would not preclude such mining done for scientific research that was publicly available without a cost. This decision is not res judicata yet and subject to debate.

With AI having the capability to generate images with unprecedented quality, it has also brought to the fore a unique issue within the larger ambit of infringement – deepfakes. Thus far, legislations have recognized impersonation as an offence under penal, privacy and information technology laws. However, in a landmark development, courts have recognized the personality of celebrities as being monetizable assets which are prejudiced by the emergence of deepfakes along with causing disrepute to their individual personas.

Rights in output data – AI-generated works, AI-inventions and other AI-outputs and infringement

In most countries, authorship of creative works and invention of new technology can only be attributed to humans and can be procured by corporations via a work-for-hire arrangement. A vital question is whether AI can be regarded as a legitimate author of the content it generates or as an inventor in the case of patents, given the lack of legal personality of the AI itself.

Pertinently, the Commission for Intellectual Property and Companies in Africa was the first global office to have granted a patent application where AI was the inventor. This move had received considerable backlash from other countries. However, the Hong Kong government has declared AI-generated works as being capable of copyright protection under the existing law. The US Patent Office has also issued a nuanced Inventorship Guidance providing a framework for examiners of patent applications to assess the quantum of human contribution for the invention to qualify for patent protection – a move seeking to balance IP rights with the need to leverage upcoming technology. In the UK as well, the law specifically permits copyright protection in "computer-generated works" though the broader question of originality being a precondition for IP protection continues to be ambiguous. In Europe, AI cannot be stated as inventor of a patent.  

Additional contributors include: Joel Bock (US), Michael Franzinger (US), Sunita Kaur Chima (Malaysia), Jennifer Cass (UK), David Wagget (UK), Aliya Seitova (Kazakhstan), Jenni Rutter (New Zealand), Nadia Ormiston (New Zealand), Güneş Haksever (New Zealand), Davin Olen (South Africa), Shahid Sulaiman (South Africa), Catherine Lee (Singapore), Andre Rahadian (Indonesia), Minh Tran (Vietnam), Linh Tran (Vietnam), Richard Keady (Hong Kong), Julian Ng (Hong Kong) and Dong-Hwan Kim (South Korea)

Download the full global AI trends report or navigate to our key areas listed below. Visit Dentons' AI: Global Solutions Hub for the latest legal insights, webinar recordings and regulatory overviews from around the world.

Download the report

• Our overview for 2025
• AI regulation, governance and ethics  
• Data privacy and cybersecurity
• AI projects and procurement
• Employment and talent management
• Disputes and managing liability
• M&A and investments
• Competition and antitrust