The Information Commission's Office (ICO) has opened a six-week call for evidence on how data protection law should apply to agentic AI, seeking views from developers, deployers and legal and technical experts on security, transparency, accountability, automated decision-making, fairness and lawful processing before it publishes guidance on the technology.
The call for evidence opened on 8 October and closes at the end of 20 November 2026. The ICO said the evidence gathered will inform future guidance giving organisations greater clarity on their obligations, and will also feed into its forthcoming statutory code of practice on AI and automated decision-making.
The ICO launched the call alongside a report on its two-year supervision programme for foundation model developers, in which it said ten developers operating in the UK, including Amazon, Anthropic, Apple, Google, Meta, Microsoft and OpenAI, had made or committed to data protection changes including clearer transparency information, stronger mechanisms for people to exercise their rights and tougher assessments of safeguards.
The regulator also confirmed it had made enquiries with OpenAI, Anthropic, Meta and the UK's AI Security Institute about recent agentic AI testing and deployment, after reports that certain agents bypassed protections, used unauthorised communication channels and accessed external systems such as Hugging Face. The ICO said it had contacted several developers and their testing partners to establish what risk assessments and safeguards were in place, and that the enquiries are ongoing.
Richard Nevinson, Director of Technology Regulation at the ICO, said the recent reports showed both how fast agentic systems were advancing and the risks they posed if guardrails were not fit for purpose. "Our message is clear: the fact AI agents act with autonomy is not an excuse for poor compliance. If people are to trust AI innovation, they rightly expect to know how their personal information is being protected," Nevinson said.
The call for evidence is structured around six themes:
1. Data security: how security issues appear within agentic AI and how organisations can address them
2. Transparency in complex agentic AI supply chains
3. Allocating and evidencing accountability across agentic AI systems
4. Understanding where the UK GDPR automated decision-making provisions apply to agentic AI and applying appropriate safeguards
5. Fairness and purpose limitation
6. Lawful processing of personal information in agentic AI
The ICO defines an AI agent as a system based on large language models that has access to tools, memory and an environment, and some autonomy in how it approaches a task, and an agentic AI system as any computing system that includes one or more agents.
It describes the emerging ecosystem in three layers: a service and integration layer covering how organisations develop, provide and obtain agentic services; an agentic AI system layer covering the components that let agents receive instructions, access data and use tools; and the AI agent layer covering how an individual agent is set up, trained and evaluated.
The ICO said the boundaries of accountability between the layers are not always clear, that organisations should assess their roles by reference to the decisions they make about processing, and that they should not rely on how they describe themselves in contracts.
On accountability, the ICO said some agentic AI developers suggest they are controllers only for initial development and for maintenance and security processing. It said organisations should interpret the concept of controller broadly, and that any organisation which determines or exerts sufficient influence over the purposes and means of processing carried out through an agentic system can be a controller for it.
Decisions about which goals an agent pursues, what personal data it can access, what models, tools and protocols it can use, what permissions it has and what actions it can take may all bear on controllership, with protocols, permissions and access controls likely to influence the means of processing more heavily than in generative AI.
The ICO is seeking evidence on how organisations allocate controller, joint controller and processor roles in practice and in contracts, what logging, tracing and monitoring mechanisms they use to track agent actions and data access, and what information developers provide to deployers about how systems work and what risks may emerge in use.
The ICO's existing guidance on AI and data protection, explaining decisions made with AI, generative AI and foundation models continues to apply to agentic workflows. The Data (Use and Access) Act 2025 reformed Article 22 of the UK GDPR to permit solely automated decisions with significant effects on the basis of most personal data subject to safeguards, including the right to human intervention, while retaining tighter restrictions where special category data is involved. The ICO said its agentic AI guidance would be kept under review as the technology develops.
The Commission said agentic systems may process personal data at several points in completing a task, including collecting it from internal and external sources, generating new information about people, sharing it with other agents, tools or third parties, and using it to make decisions or take actions.
Because an agent may select different tools, sources and recipients each time it performs the same task, the ICO said that without appropriate safeguards an agentic system could use more personal data than necessary or disclose special category data to an unauthorised third party, and that orchestration across multiple agents could allow inaccurate information, inferences or bias to spread quickly through the processing chain.
The call for evidence builds on the ICO's input to a Digital Regulation Cooperation Forum deep dive on agentic AI in summer 2025, the DRCF's call for views and March 2026 foresight paper, and the ICO's own Tech Futures report on agentic AI published in January. The ICO is separately examining the use of agentic AI in the adtech ecosystem and expects to publish a report on that later this year.
Responses may be published in full or in summary, and the ICO said it would consider any freedom of information request for responses in accordance with the legislation.
Agentic AI call for evidence: https://ico.org.uk/about-the-ico/ico-and-stakeholder-consultations/2026/10/agentic-ai-call-for-evidence/
