Ofcom Publishes Strategic Approach to AI

Ofcom has published its strategic approach to AI 2024/2025 setting out how the communications regulator plans to approach AI in the sectors it regulates, including fixed and mobile telecoms, TV and radio broadcasting, on-demand services, video sharing platforms and online safety. 

Ofcom’s publication of this strategic approach follows a direct request in the UK government’s consultation response to the March 2023 AI  White Paper (see our blog here) for key regulators to publish their strategic approaches to AI by the end of April 2024. This ask was repeated in letters sent directly to these regulators by ministers on 1 February this year.  

Ofcom’s Strategic Approach and work to date

AI principles

In its strategic approach, Ofcom emphasises its support for the Government’s 5 AI principles which cover: (1) safety; (2) security and robustness; (3) appropriate transparency and explainability; (4) fairness accountability; and (5) governance; contestability and redress. It considers that they are broadly aligned with the underlying principles of Ofcom’s regulatory regimes and the outcomes it wants to see across its sectors.

The principles are not currently on a statutory footing, although the UK government has said it will keep this under review, and use the regulator responses (including this response from Ofcom) to help inform its decision making around this. 

Update on activities

Ofcom’s publication recognises the potential benefits of AI whilst also acknowledging possible downsides. These include the creation and spread of illegal or harmful content. It also provides an update on its activities relating to its:

• regulation of services that use AI technologies;
• work to date to understand and tackle AI risks;
• capability to address AI risks;
• cooperating with others on AI issues; and
• planned AI work for 2024/2025.

Current areas of focus

In detailing some of the steps Ofcom has already taken to address AI related risks, it is evident there has been a focus on three “cross-cutting” areas - the issues of synthetic media, personalisation, and security and resilience. 

One step taken by Ofcom to tackle the first two of these is to publish a draft Illegal Harms Code of Practice under the online safety regime. This includes proposed measures that relate to accountability and governance to identify and manage risks, including risks posed by the sharing of illegal synthetic content. It also recommends that certain services collect safety metrics when testing recommender systems (which may include AI-driven systems), to improve online safety. In relation to the third area - security and resilience - the actions taken include tracking developments in how GenAI could be used to develop malicious tools that threaten network security. 

Next steps for 2024/2025

Ofcom’s plan for 2024/2025 is to continue identifying and responding to AI related risks across its remit, to take forward the key activities already undertaken to date (such as those summarised above) and to start new work across different policy areas such as online safety. For example, it plans to draw up new online safety codes of practice; to research the merits of synthetic media detection tools and automated content classifiers, and to research the merits of using GenAI for content moderation. It is also intending to consult on its information gathering powers, including powers it may use to gather information about algorithms, and to monitor international regulatory developments including the EU’s AI Act. 

Interestingly, the last item on Ofcom’s list of planned work for 2024/2025 is to continue building its AI capabilities by upskilling its talent. Like many regulators, it is also exploring how it can leverage AI across its own operations. 

For more information on the approaches taken by other key regulators, see our blogs on the strategic approaches to AI published by the CMA here, the FCA here and the ICO here. A centralised resource listing all strategic approaches published has been made available by the government here.