UK’s House of Lords to debate AI regulations
Discussion will explore the need for effective AI rules in light of risks | UK government’s IP plans may be scrutinised | HoL publishes key report on country’s AI strategy.
The UK’s House of Lords (HoL) will next week debate how AI should be regulated, with the government’s proposed IP reforms likely to feature.
The second chamber of Parliament will discuss a motion tabled by Lord Ravensdale that “this House takes note of the ongoing development of advanced artificial intelligence, associated risks and potential approaches to regulation within the UK and internationally,” on 24 July.
In advance of the debate, the HoL released an in-depth report yesterday (July 18) titled Artificial Intelligence: development, risks and regulations, in which it acknowledged the rapid growth of AI, from “generative language models such as OpenAI’s ChatGPT to advances made in medical screening technology.”
The report highlights the UK government’s sentiments towards AI as outlined in its National AI Strategy: that AI is the fastest growing deep technology in the world, with “huge potential to rewrite the rules of entire industries, drive substantial economic growth and transform all areas of life”.
However, the HoL report also notes the AI’s risks, including legal and regulatory challenges, bias and discrimination, privacy, ethical, and security concerns.
Monday’s debate will discuss how to best regulate AI and the differences in approaches that exist internationally.
HoL wins concession on data mining copyright exception
The HoL communications and digital committee examined AI’s impact on the creative industries in January, 2023.
It outlined a concern about AI tech, particularly text and data mining, which uses existing materials to train AI models.
The committee also criticised the proposed IP law reforms, specifically a copyright exception for text and data mining, labelling them as “misguided and harmful”.
It has urged the Intellectual Property Office (IPO) to pause the changes and conduct an impact assessment.
The committee also cautioned against the use of AI to generate and distribute creative works, stressing the need to protect the rights of performers and original creators.
In response to the committee's concerns, the government announced in April that it would not proceed with the proposed exception for text and data mining of copyrighted works.
Instead, it pledged to collaborate with users and rights holders to develop a "code of practice" on text and data mining by AI, expected to be published by summer 2023.
UK’s proposed AI regulatory framework
The UK’s strategy for regulating AI aims to balance promoting innovation and responsible governance.
The strategy, outlined in a white paper following the National AI Strategy, emphasises long-term investment and supporting the transition to an AI-enabled economy.
The government acknowledged the need to fill regulatory gaps and mitigate emerging challenges such as AI discrimination.
To address these concerns, the UK government proposed a regulatory framework based on five core principles: safety, security, and robustness; transparency and explainability; fairness; accountability and governance; and contestability and redress.
Initially, the principles will be implemented by existing regulators, with a future plan for statutory duty, according to the HoL report.
EU and US comparison
In contrast to the UK's approach, the EU is taking a more comprehensive approach to AI regulation by proposing a set of rules applicable to all sectors, focusing on risk levels associated with different AI applications.
The EU AI Act establishes categories of risk and corresponding regulations for AI systems.
Meanwhile, the US has adopted a risk-based and sector-specific approach to AI regulation, distributed among various federal agencies.
However, the US “have not created an even or consistent federal approach to AI risks”, said the HoL report.
Did you enjoy reading this story? Sign up to our free daily newsletters and get stories sent like this straight to your inbox
