Artificial intelligence | UK Regulatory Outlook March 2024

EU updates

EU AI Act adopted by Parliament

The EU AI Act was formally adopted by the European Parliament on 13 March 2024 – see the adopted text. This is not the last step in the legislative process but is an important milestone.

The speed at which the texts were finalised for this vote means that final checks and corrections are outstanding. A corrigendum is expected, and scheduled to be accepted by a quick Parliament vote on 11 or 12 April.

The next step after the Parliament's second vote will be formal adoption of the text by the Council of the EU. We have not yet seen a date for this vote, but it is expected in April. The Act will then be published in the Official Journal of the EU and enter into force twenty days after – expected to be late May.

UK updates

ICO launches second consultation on generative AI

The UK Information Commissioner's Office (ICO) has launched a second call for evidence in its series of consultations on generative AI. This consultation looks at "Purpose limitation in the generative AI lifecycle" and is focused on how the data protection principle of purpose limitation should be applied at different stages in the generative AI lifecycle.

The call for evidence closes on 12 April 2024 and the responses can be submitted here.

The first consultation focused on the lawful basis for web scraping to train generative AI models and closed on 1 March 2024.

International updates

Council of Europe reaches agreement on convention on AI

After a final week of talks, the Council of Europe announced that political agreement has been reached on the Artificial Intelligence, Human Rights, Democracy and the Rule of Law Framework Convention.

However, as many feared, this is not a convention with universal scope and application. It has been reported that the treaty will only apply to public authorities or private actors acting on their behalf. 

Council of Europe member countries will vote on the text in May, after which signatory countries will need to ratify it at national level.

Separately, the European Data Protection Supervisor (EDPS) published a statement on the negotiations around the convention. The EDPS is concerned that the convention could be "a missed opportunity to lay down a strong and effective legal framework for the development and uptake of trustworthy AI." Its concerns relate to: (1) the very high level of generality of the legal provisions of the treaty; (2) its largely declarative nature which would lead to different applications across signatories; (3) limitation of the scope only to public authorities; and (4) the lack of "red lines" and criteria for banning AI applications that pose unacceptable threat.

G7 discussions on AI

The G7 countries (Canada, France, Germany, Italy, Japan, UK and USA, with the EU also participating) issued a Ministerial Declaration on 15 March at the end of a two-day summit on digital and technology matters. AI-related matters in the communiqué included:

• developing tools and mechanisms to hold accountable businesses that commit to their non-binding code of conduct (issued as part of the G7 Hiroshima AI process);
• addressing the topics of how to ensure continuing effective competition in AI markets at the next G7 summit, scheduled for later this year in Rome;
• developing a toolkit for "the safe, secure, and trustworthy development, deployment, and use of AI in the public sector";
• producing a report on the "factors and challenges of AI adoption" by the private sector, particularly small and medium businesses, to inform policymaking and initiatives to support adoption.

OECD explanatory memorandum on the definition of AI system

The Organisation for Economic Co-operation and Development has published an explanatory memorandum on its updated definition of AI. It has also published an accompanying blog post explaining the main points. This definition formed the basis for the definition of AI under the EU's AI Act.

And finally …

We have updated our overview of the risks to be considered by a business using AI to reflect the current position with the AI Act and in the UK.