Skip to content

Artificial Intelligence and Intellectual Property: what next for the UK?

Author
Robert Dickens

Trainee

London

View profile →

16 December 2021

There is still time to provide input to the UKIPO on copyright protection of computer-generated works, copyright exceptions for text and data mining and AI inventorship of patents as the government bids to make Britain a global AI superpower within the next ten years.

The UK Intellectual Property Office is conducting a consultation on artificial intelligence and intellectual property (the Consultation). In light of the Government's call for views on the interaction between AI and IP (March 2021), the consultation looks to help deliver on the long-term aims set out in the UK Innovation Strategy (July 2021) and the National AI Strategy (September 2021). There is still time to contribute to the consultation as the closing date for submissions is 07 January 2022. The government will then publish a response document in due course.

The aim of the Consultation is to consider possible changes to IP legislation that will encourage innovation in AI technology and promote its use for the public good, while preserving the central role of intellectual property in promoting human creativity and innovation.

The Consultation focuses on three key areas: (i) copyright protection for computer-generated works without a human author; (ii) licensing or exceptions to copyright protection of information where that information is used for the purposes of text and data mining; and (iii) patent protection for AI-devised inventions.

Copyright protection for computer-generated works

In the UK, copyright currently provides protection to original literary, dramatic, musical and artistic works generated by a computer, where there is no human creator, for a period of 50 years from the date the work is made.
 
The UK position is unusual and has been criticised on the basis that copyright (indeed, any IP right) results in a cost to third parties and that computers do not need a reward or an incentive to produce original content, so there is no justification for the cost imposed on third parties. The UK position has also been criticised on the basis that the legal test of "originality" cannot be readily applied to computers because it is based on the human characteristics of personality, judgement and skill.
 
In light of this criticism, the Consultation notes that the government has explored three options:
  1. Maintenance of the status quo if it can be shown that, while machines do not need an incentive to produce original content, the current UK position encourages new AI-generated works and investment in AI.  
  2. Wholesale removal of the current protections if copyright is not needed to incentivise the use of AI or that the granting of copyright to AI results in an unreasonable cost to third parties.  
  3. A middle way whereby the duration of protection granted to an original work reflects the amount of human effort or investment put into the work.  The Consultation suggests that 5 years could be considered.  The aim of this middle way is to balance the incentive to use and invest in AI with the rights of third parties.

The Consultation does not propose any changes to the current law that protects AI-generated broadcasts, films, sound recordings and published editions. This is on the basis that these "entrepreneurial works" do not need to be original and (consequently) the rights attached to these works are narrower and shorter than those granted to original works.

Copyright exceptions for text and data mining

Text and data mining is defined in the Consultation as "the use of automated computational techniques to analyse large amounts of information to identify patterns, trends and other useful information."  

The current copyright position on text and data mining was introduced in 2014. The UK permits the making of copies of any copyright work for non-commercial scientific research by text and data mining, provided the researchers have lawful access to view that work (for example, by way of subscription) and acknowledge the work and the relevant rights holders where possible.  This copyright exception does not apply to databases.

Text and data mining of copyrighted works is also permissible under the temporary copying exception (provided the copy has no independent economic significance) and under licensing arrangements (for example where scientific research by text and data mining is taking place on a database and/or for a commercial purpose).

The main criticism of the above position is not in relation to the legal framework but, instead, a point of policy. AI systems need to be trained via text and data mining. To the extent that commercial investment in AI is taking place, or that any AI training is taking place on databases, the current copyright exception does not apply and copyright licences must be in place. In view of the National AI Strategy to make Britain a global AI superpower in the next ten years, the question is whether this exception needs to be broader to allow for greater progress and to promote new discoveries in the AI sphere.

The Consultation notes that five policy options are being considered:

  1. Maintenance of the status quo if it is shown that the current law is not hindering access to material for text and data mining, particularly in the context of training AI systems.
  2. Improving the licensing environment by creating model licences and codes of practice for copyright licensing in the context of text and data mining.
  3. Extending the copyright exception to cover commercial research and databases.
  4. A new exception for text and data mining conducted for any purpose by anyone with an opt-out system for rights holders to carve out their works.  Such an exception would not be restricted to scientific research and could include, for example, journalism or business analytics. Although, researchers would still need lawful access to read the copyrighted works and databases in the first instance. This approach, it is suggested, would remove the high costs associated with mining works for which it is difficult to locate the relevant rights holders while allowing rights holders to license their works if they wish.
  5. A new exception for text and data mining conducted for any purpose by anyone without an opt-out system. Although, as above, researchers would still need lawful access to read the copyrighted works and databases, rights holders wouldn’t be able to prevent text and data mining of these works and databases.

Patent protection for AI-devised inventions

The third and final topic of the Consultation is highly topical.  We have previously reported on the recent case law relating to machine inventorship in the context of patents, including Thaler v Commissioner of Patents [2021] FCA 879 in Australia and Thaler v Comptroller General of Patents Trade Marks And Designs [2021] EWCA Civ 1374 in the UK.

The current position in UK patent law, set out by the Court of Appeal in Thaler, is that the human(s) using AI to devise inventions should be named as the inventor in most cases.  This approach retains the protection for, and incentive to create, such inventions. However, the Consultation argues that, in line with the government's overarching policy of promoting investment in AI in Britain, this rule could be improved to support innovation as we near the point where AI-devised inventions may not require any human involvement.

 
The biggest concern set out in the Consultation is that, if AI-devised inventions cannot be protected by patents, then there may be less investment in AI and an uptick in the use of trade secrets – either of which would hinder innovation.  On the flipside, it is noted that patent protection for AI-devised inventions may result in a handful of parties with the best AI technology and training data quickly acquiring a substantial number of patents. This would damage both innovation and competition. 
 
It is notable that only a small number of countries have granted patents where an AI system is the inventor. The vast majority of countries, including the five territories in which 80% of the world’s patent applications are filed, only permit natural persons to be named as inventors.
 
Four policy options are contemplated in the Consultation:
  1. Maintenance of the status quo such that patents are only available where the inventor is human.
  2. Expanding "inventor" to include the humans responsible for the AI system that devises the invention. This is a minimal change in law whereby the person(s) responsible for programming, configuring, operating (etc.) the AI system would be considered the inventors of the patent devised by that AI system. This option would not require the human named to disclose whether or not AI was used to devise the invention.
  3. Allowing patent applications to identify AI as the inventor by either allowing AI to be named as the inventor or by removing the requirement to name an inventor if the invention was devised by AI.  In both cases, the human responsive for the AI system would be entitled to the patent rights in the first instance (these rights passing to a legal person such as a corporate entity by virtue of contractual rights where applicable).  AI would not be given the right to apply for or own patent rights. The Consultation considers that the disclosure of whether an invention was devised by AI or not might stimulate use of and investment in AI.
  4. Protecting AI-devised inventions with a new kind of right. The Consultation envisages a quasi-patent right with limited exclusive rights and a shorter duration to operate alongside the current patent system. The Consultation also puts forward suggestions for a legal framework for these AI quasi-patents based on the current regime, including a stricter test of inventive step, as AI may invent in ways that humans would not deem obvious.  Alternatively, that there should be no test for obviousness such that novel AI-devised inventions are automatically protected. Depending on the circumstances, a new quasi-patent right might be less attractive than a patent if it provides less protection or, alternatively, might be more attractive if the legal hurdles are easier to overcome. The Consultation notes this issue and the foreseeable consequence that inventors seeking protection of their innovations might falsely acknowledge the use of AI or not dependent on their chances of success within each regime.

Conclusion

These are no doubt exciting times. Will the government stick to the status quo or will it break the mould and deliver an enhanced legislative agenda in pursuit of its national strategy to make Britain an AI superpower? Either way, it is clear that the increasing pace of technological development and profound leaps in AI capability will continue to challenge existing IP concepts and principles over the next few years.
 

Related blog topics