As AI solutions continue to become more commonplace, what to do about IP rights for AI developments and deployments are becoming more pressing concerns. Difficult issues are being debated by international organisations such as the World Intellectual Property Organisation (WIPO) (see post here). This time, it is the European Parliament’s turn. In April 2020, the EP published a draft report on IP rights for the development and deployment of AI technologies (see post here). In May 2020, the report received a number of suggested amendments, which shine light on the key pressure points and concerns around AI regulation.
For those new to these reports, they relate to the protection of IP rights in relation AI technologies. In particular, they look at the patentability of AI innovations, and the challenges this poses in light of the complexities of certain AI technologies, as well as data protection and issues around the autonomisation of decision-making processes that can give rise to AI-generated technical or artistic creations.
Certain key and controversial themes emerged from the proposed amendments to the report, with the most notable being:
- The amendments generally spoke to the difficulty in striking a balance between sufficiently regulating AI technologies and protecting investment on the one hand, and incentivising creation and sharing on the other. For instance some amendments highlighted a desire to remove unnecessary legal barriers to the development and use of AI technologies, noting that policy choices in this context should be flexible so as to not hamper growth or innovation, and recognising the importance of data sharing and open source technology. By contrast, other amendments highlighted that AI technologies should not come at the expense of the interests of human creators whose copyright work may be used during the development, deployment, training and use of AI-enabled systems, recommending that efforts to encourage the use of AI technologies should be limited to those who comply with the Union’s ethical principles and regulatory frameworks.
- On patentability, some amendments indicated that this could serve to incentivise innovation, whereas others suggested that the threshold of patentability for AI inventions should be lowered in relation to socially beneficial areas, such as healthcare, the environment, criminal justice, culture and education.
- The distinction between human-generated works, and the protection thereof, and AI-generated works was a particularly controversial theme. Amendments called for IP rights over the development of AI technologies and IP rights over AI-generated content to be kept separate, emphasising the notion that originality is directly linked to natural persons. One amendment also made the point that copyright protection over certain works that had been generated with the assistance of AI technology could result in a monopoly of a handful of businesses.
- On the subject of data, some amendments noted that copyright-protected works used as input to train AI processes do not require any new policy consideration and should remain under the current copyright/regulatory framework. Other amendments stressed that a balance must be struck between the promotion of wider use and sharing of data, and the protection of IP rights, privacy and trade secrets. One amendment also made the point that unnecessary barriers to data in the development of AI technologies could result in selection bias for training data, which could in turn result in more harmful bias in results - increased flexibility in the use of protected data was therefore required.
It remains to be seen how such suggestions and concerns might be reflected in legislation going forward - however they provide a valuable early insight into those areas that will be most contested as regulators attempt to strike the balance between providing sufficient protection of rights, while nevertheless allowing the flexibility required for innovation to flourish.
Thank you very much to Anna Bashall for her research in preparing this post.