UK ICO Publishes Report on Data Protection Expectations for AI-Enabled Online Advertising
Time 4 Minute Read

On November 13, 2026, the UK Information Commissioner’s Office (“ICO”) published a report examining how large language models (“LLMs”) and agentic artificial intelligence (“AI”) are reshaping online advertising. The report considers how existing requirements under the UK General Data Protection Regulation (“UK GDPR”), the Data Protection Act 2018, and the Privacy and Electronic Communications Regulations 2003 (“PECR”) apply to these emerging technologies. The ICO’s general message is that LLMs and agentic AI do not create a new set of data protection requirements for online advertising. They may, however, change how organizations need to apply those requirements and demonstrate compliance, particularly as systems generate richer inferences and undertake advertising activities with greater autonomy. Below is a summary of certain key points discussed in the report.

The report reviews several developing use cases. These include advertising placed alongside chatbot responses, natural-language interfaces used to manage advertising campaigns, AI agents that coordinate advertising activities across multiple service providers, and AI-enabled advertising in retail media, product discovery, and connected television. The ICO distinguishes between practices that are already in use, those still being tested, and those that remain more conceptual. It also makes clear that the report is focused on activities involving storage or access technologies, personal data processing, or both, rather than on AI tools used only to generate advertising content (i.e., visuals).

The report emphasizes that organizations remain responsible for their data protection obligations even where AI agents make intermediate decisions or automate campaign activity. Whether a party is acting as a controller, joint controller or processor depends on its actual role in the processing, not on contractual labels alone. The ICO also notes that an advertiser will not necessarily be responsible for every processing activity simply because it sets the overall campaign objective.

The ICO also indicates that meaningful human oversight requires more than reviewing campaign results after the fact. Organizations should understand the objectives assigned to an AI system, the categories of personal data the system can access, the actions it is permitted to take, and the stages at which human review or intervention is possible. The report further notes that the presence of some human involvement does not necessarily mean that a decision falls outside the UK GDPR’s rules on solely automated decision-making.

With regards chatbot interactions and inferred audience traits, the ICO notes that conversational inputs may reveal sensitive information about an individual’s circumstances, interests, or intentions, and that systems may infer characteristics the individual did not expressly disclose. According to the ICO, inferred information may qualify as special category data where it is used to infer or act on a sensitive characteristic, even if the inference is uncertain. For example, if a system infers a possible health condition from a chatbot exchange and uses that inference for ad targeting, the ICO considers that processing to involve special category data. The report reiterates the ICO’s position that using special category data for online advertising requires explicit consent.

Advertising and the underlying data processing must be transparent. In this respect, users should be able to distinguish paid advertising from generated responses or recommendations and should understand how their information influences ad selection. The ICO notes that broad statements that a service uses AI to personalize advertising will not be enough if they do not explain what data is used and which organizations are involved.

The ICO states that organizations should not rely on high-level vendor claims without examining the underlying practices. Descriptions such as “agentic,” “contextual,” or “privacy-preserving” do not, in the ICO’s view, establish how a system actually works or whether it complies with applicable law. Instead, organizations should seek evidence regarding data sources, processing flows, the scope of an agent’s authority, and any limits on the information that can be exchanged across systems.

Lastly, the ICO reiterates that PECR applies where technologies store information on, or access information stored on, a user’s device for online advertising purposes. In the ICO’s view, that consent requirement covers not only ad selection and delivery, but also related tracking and profiling activities. Where PECR consent is required, the ICO’s view is that consent will also be the appropriate lawful basis under the UK GDPR for related personal data processing, provided it is valid and sufficiently specific. The report also stresses that automated workflows should be able to communicate and honor consent withdrawals across participants.

The report findings will inform the ICO’s forthcoming work on agentic AI and automated decision-making, including its AI and automated decision-making code of practice. Read the report.

Search

Subscribe Arrow

Recent Posts

Categories

Tags

Archives

Jump to Page