Artificial intelligence | UK Regulatory Outlook June 2024

UK updates

General election: AI in the manifestos of the two main political parties

The two main parties fighting to win the UK general election on 4 July 2024 have published their manifestos. The UK has so far taken a light-touch approach to AI regulation and simply asked existing regulators to develop an AI strategy based on their existing powers, shaped by non-statutory "high level principles".

Our Insight explores in detail whether we should expect change after the general election based on the parties' views shared so far. Key points are:

• The Labour Party would legislate to put voluntary codes on AI safety onto a statutory footing (but timeframes are unclear).
• A Labour government would also discuss with stakeholders whether new legislation on AI in the workplace is needed.
• Both main parties would legislate to outlaw sexualised/sexually explicit deepfakes.
• Both would drive greater use of AI in the healthcare sector specifically and in government/the public sector more generally.
• The Labour party will reform planning to make it easier to build data centres and will create a "National Data Library" which could both support AI development in the UK.
• There is no reference in the Labour manifesto to the tension between IP rights and using web-scraped data for AI training. This is a difficult issue that the current Conservative government has been trying to resolve for some time, which the Conservative manifesto commits to continue.
• Neither party shows any sign of wanting to create a new overarching horizontal regulatory framework for AI along the lines of the EU AI Act.

Our Insight on the manifesto pledges also discusses the recommendations of the UK House of Commons Science, Innovation and Technology Committee in its report into the governance of AI. The inquiry was launched in October 2022, with an interim report in August 2023 outlining twelve challenges which AI governance should address (see this Regulatory Outlook). The government published its response to the report in November 2023. The latest report updates those recommendations to "apply to whoever is in Government after the general election."

EU updates

AI Act will not be law before August

EU officials have confirmed to us that the definitive text of the EU AI Act will not be published in the Official Journal of the EU until mid-July at the earliest, due to its length and a significant backlog. It will come into legal effect 20 days later – now delayed into August 2024.

The first compliance deadlines (for the prohibitions) will now be some time in February 2025, rather than before the end of the year as had been expected.

Even though the first deadline has been delayed, businesses which will be potentially affected still need to plan and prepare for it. Our webinar on 28 June will cover the AI Act's implications for various business sectors, explaining the Act's framework and the phased deadlines for compliance.

Sign up here: Get ready for the EU AI Act – first deadline in six months and see our Insight.

EU AI Office starts its work

On 29 May 2024, the EU Commission announced the establishment of the AI Office which will play a key role in the implementation of the AI Act, working with national-level enforcement bodies. See this Regulatory Outlook for more background. The office has already begun its work. 140 employees are split between technology specialists, administrative assistants, lawyers, policy specialists and economists.

The AI Office comprises five units:

• a Regulation and Compliance Unit that will ensure the consistency of the regulatory approach in the application and enforcement of the Act across the EU;
• an AI Safety Unit with a focus on the identification of systemic risks of very capable general-purpose models, ways to mitigate their risks and evaluation and testing approaches;
• an Excellence in AI and Robotics Unit which will support and fund research and development;
• an AI for Societal Good Unit which will ensure the Office's engagement in international initiatives for good, such as weather modelling, cancer diagnoses and digital twins for reconstruction; and
• an AI Innovation and Policy Coordination Unit to oversee the implementation of the EU AI Strategy.

The AI Office is preparing guidelines on the definition of "AI system" and on the prohibitions on certain types of AI system, which will take effect six months after the entry into force of the AI Act. The office will also develop codes of practice in relation to the AI Act provisions on general-purpose AI models, which the AI Act requires (Article 56(9)) to be ready no later than nine months after entry into force of the Act.

French CNIL consults on GDPR application to AI system development

The French data protection authority, la Commission nationale de l'informatique et des libertés (CNIL), has launched a further consultation on its AI "how-to sheets".

The CNIL has previously consulted on its "how-to sheets" for AI development – see this Regulatory Outlook for background. The CNIL is now consulting on the legitimate interest as a legal basis for development of AI systems and in relation to the dissemination of open source models and web scraping; informing data subjects and respecting and facilitating the exercise of the rights of data subjects; data annotation; and ensuring the safety of AI system development.

The consultation closes on 1 September 2024. Alongside this consultation, the CNIL has published a questionnaire on the application of the GDPR to AI models which explores when AI models can be considered anonymous and when they are regulated by the GDPR where personal data is used for training purposes.

Previously, the CNIL published its first set of AI "how-to sheets" to help organisations develop AI systems in compliance with the GDPR. The recommendations also take into account the AI Act. The scope of how-to sheets is limited to the development phase of AI systems, not deployment, which involve the processing of personal data, for example during training of AI systems. The recommendations are illustrated by examples and case studies.

International updates

G7 leaders' AI plans

Following the G7 summit held on 13-15 June 2024 in Apulia, Italy, the leaders of the G7 nations (Canada, France, Germany, Italy, Japan, UK and USA, as well as the EU) have published a final communiqué setting out their plans for mutual work.

In relation to AI, the G7 leaders outlined their plans to launch an action plan "on the use of AI in the world of work" which they ask the relevant ministers for social and employment matters to develop. The G7 countries also call on their competition authorities to monitor the development of the AI industry to address potential competition issues and prevent adverse effects at the outset. Finally, the leaders pledged to develop a brand to support the implementation of the International Code of Conduct for Organisations Developing Advanced AI Systems announced as part of their Hiroshima AI Process (see this Regulatory Outlook for background).