UK wants to boost AI development by removing barriers to data mining – TechCrunch
The UK is considering changing an existing law to allow text and data mining “for any purpose”, in a bid to boost the development of artificial intelligence (AI) across the country.
The announcement is part of a wider strategy to ‘improve’ AI and transform the UK into what it calls a ‘global AI superpower’ – and part of that will involve reassessing existing laws on intellectual property (IP). After two months consulting period where stakeholders from across the industry spectrum were invited to provide input, including rights holders, academics, lawyers, trade organizations and companies, the Intellectual Property Office (IPO) of the UK today published their response and confirmed what will (and won’t) change going forward.
Text and data mining (TDM) is essential to the development of new AI applications, allowing researchers and companies to copy and exploit disparate data sets to train their algorithms. However, accessing enough relevant data presents inherent challenges – the data is often owned by third parties who may only want to make the data available under a commercial license, if they make it available at all.
Return in 2014the UK has amended its existing TDM regulations, which are linked to the Copyright, Designs and Patents Act (1988) — to include an “exception” allowing AI researchers to exploit third-party data for non-commercial purposes without incurring significant costs. However, this still imposed significant limits on how the data could be used and discouraged companies from investing in AI development. Moreover, it did not extend to database rightswhich differs from works covered by traditional copyright.
Today’s announcement essentially solves that problem. The UK government is now considering adopting a TDM exception that covers any purpose well beyond research and academia, with no opt-out for rights holders. In addition to that, it also includes provisions for database rights.
This goes against the comparable European Union (EU) Directive on copyright in the digital single market, which provides a mandatory exception only for TDM in the field of scientific research. Indeed, rights holders can remove their copyrighted work from commercial use cases, which means they can still monetize TDM.
The UK’s proposed changes could be an important part of its stated “upgrading” plans, given that access to AI training data is a major stumbling block for all businesses, except the biggest ones. But more than that, it seems the intention here is to attract AI companies to the UK, knowing that they have more freedom to perform text and data mining. This is particularly important in a UK which is now in direct competition with the EU after exiting the bloc in 2020, as he acknowledges in the response published today:
These changes make the most of the increased flexibilities resulting from Brexit. They will help make the UK more competitive as a location for data mining companies.
The Brexit effect
This new exception to copyrights and database rights, which the government plans to enshrine in “appropriate legislation” in due course, effectively shifts the balance of power from rights holders to corporations and other entities. commercial. But the transition could have unintended consequences, some say. Under the proposed new rules, the data miner must still acquire the data through legal means, which means that it must be publicly available (e.g. through a subscription). As a result, rightsholders who could previously charge for TDM under a data licensing model may instead withhold their data completely, which could negatively impact the development of AI in the future. .
“On the one hand, this decision can be seen as a catalyst for the development of AI, but on the other hand, it may have the effect of encouraging copyright owners to impose more restrictions on their content,” said Richard Johnson, partner at European intellectual property law firm Mewburn Ellis. , TechCrunch said in a statement.
It should be noted that rightsholders themselves will still have certain rights under the new rules, in terms of where they choose to post their data or copyrighted works, and they can always charge for access to this data. They “will no longer be able to charge for UK licenses for TDM” in particular, and they will not be able to apply any type of opt-out – any entity that legally accesses the data will be able to exploit it.
The response published today by the UK Government is remarkable not only for what is changing with text and data mining, but also for what is not changing. At the heart of the consultation was whether computer-generated works (CGWs) without a human author should continue to be protected by copyright law – the UK is in fact one of the few countries to grant copyright protection to CGWs, with a 50-year protection period in place (this compares to 70 years for human-generated works).
As the UK debated whether the period of protection should be changed or removed, it finally decided not to change anything, noting that existing protections around CGWs were not “harmful” and that the use of AI is still in its infancy. “We will keep the law under review and may modify, replace or remove the protection in the future if the evidence supports it,” the response noted.
Similarly, the UK has also ruled that AI systems still cannot be patented for inventions, despite some movements in this direction in a few jurisdictions around the world. The main reasons cited are that AI is not yet advanced enough to “invent” without significant human input – a legal opinion shared by most countries. In its response to the consultation, the UK said it was reluctant to deviate too far from “international standards” on inventorship.
“This may be viewed by some as a lost opportunity – however, the commitment to actively engage in the search for international consensus is very positive to see,” Johnson said. “From an end-user perspective, it would be undesirable for the IP landscape to be fragmented on this issue. The government’s response is therefore consistent with many court rulings in this area, in that it recognizes the potential for change in the future, but avoids immediate action.