The UK’s Intellectual Property Office (IPO) has proposed a change to copyright law that would allow data scientists to mine text and data to train commercial AI and machine learning models without fear of infringing copyright or being obliged to pay additional fees.
The IPO made its proposal in response to a consultation on AI and intellectual property law, which followed the publication of the UK’s national AI strategy last September.
The consultation also considered whether AI systems can be recognised as inventors in patent applications, as some have called for, but the IPO has decided not to propose any changes to the UK’s ‘patent inventorship criteria’.
New UK copyright law for AI data mining
The government says the proposed changes to data mining rules “will help make the UK a location of choice for data mining and AI development”, supporting its ambitions to become a global leader in AI innovation and research.
Related
AI and automation
Meta’s new Sphere AI tool will check the accuracy of Wikipedia entries
AI and automation
IBM acquires Databand to improve machine learning models
AI and automation
Vodafone pens ‘risky’ AI deal with Google Cloud
AI and automation
Emotion recognition is mostly ineffective. Why are companies still investing in it?
“We want to ensure the UK continues to have one of the best intellectual property frameworks in the world. IP is key to innovation,” said science minister George Freeman. “Our new UK rules on copyright and data mining will act as a catalyst for our innovators to flourish, helping ensure the UK’s IP system remains a powerful enabler for ground-breaking R&D.”
Under the current rules, which have been in place since 2014, data scientists are allowed to access and use third-party datasets to train their AI models without running the risk of incurring costs, but only for “non-commercial” uses. The new rules would remove this limitation, meaning that anyone with legal access to a dataset could use it for commercial AI development, and the copyright holder would be unable to impose a charge for doing so.
The IPO’s proposals mark a departure from EU law, which only allows text and data mining for scientific research, and gives the owner of datasets the option to opt out or monetize their information.
Content from our partners
How clinical trials infrastructure is undergoing digital transformation
Webinar – Top 3 Ways to Build Security into DevOps
Tech sector is making progress on diversity, but advances must accelerate
Positive move for competitive AI development
The new text and data mining rules are a positive move for companies developing AI, says Ryan Abbott, a professor at the University of Surrey’s School of Law. “We have only recently had machines generating economically valuable creative works at commercially significant scale very recently, and allowing protection encourages people to develop and use AI to create useful works,” he says.
Imogen Ireland, an IP lawyer at global law firm Hogan Lovells, says the text and data mining proposals “signal that the UK government is committed to making the UK a competitive market for the development of AI.”
Rights holders, particularly those in the music industry, voiced concerns about the changes as part of the consultation, Ireland explains. “I think the government’s answer back to them is that it will eventually benefit you, too,” she says.
“The government talks about putting various protections in place for rights holders, such as being able to choose the types of platform which can use your data so you can pick one that may benefit you in future, and we need to see how the market as a whole will respond to this.”
The final shape of the legislation has yet to be determined, Ireland added. “There’ll be a period of reflection, and we’ve yet to see how this will surface as legislation.”
AI inventors: no changes to UK patent criteria
The consultation also considered the thorny issue of whether AI can be considered an inventor. At present, most jurisdictions, including the UK, EU and US, require a human inventor to be named on a patent application, meaning IP generated by machines cannot be easily patented. But a landmark case in South Africa last year saw it approve its first patent for an AI-generated invention.
The UK government has decided that “no changes will be made to the UK’s patent inventorship criteria or copyright computer-generated works provisions at this time”. But, it says, it will “keep AI technical development under review to help ensure that UK inventorship rules continue to support AI innovation and will seek to advance discussions internationally to support the UK’s economic interests.”
Abbot, who represented the developer who brought the AI invention case to court in South Africa, is disappointed the UK has declined to put measures in place to protect AI-generated inventions. “It says to industries where AI is being increasingly used in R&D, such as for drug discovery and repurposing in the pharmaceutical sciences, that companies cannot use machines to generate innovation if they need patent protection,” he argues.
But the decision not to take a stance on AI inventors is a sensible one at the current time, argues Ireland. “The law is relatively well set up to deal with the way AI is currently being used in innovation,” she says. “At the moment there are a number of use cases where AI is used to assist human innovation, and in these cases, it is generally possible to identify a human inventor.”
Ireland says making AI a “new class of inventor” could negatively impact other areas of law. “It’s important to harmonize changes across the other patent regimes,” she says. “So it wasn’t a surprise [that the government has not recognised AI inventors], and I think a lot of people will be pleased to have more time to think about this question and see how the use cases for AI evolve.”
Read more: New legislation ‘urgently needed’ on biometrics
Topics in this article: data mining, IPO, Machine Learning, Patents