UK's ICO secures data protection commitments from 10 major AI developers and moves its scrutiny to AI agents

Who the ten are, and what they promised. Per the ICO's official news page, the ten biggest foundation model developers operating in the UK are Amazon, Anthropic, Apple, Cohere, DeepSeek, Google, Meta, Microsoft, OpenAI and Stability AI. The ICO says that following its scrutiny these companies have made, or committed to make, data protection changes in three categories: clearer transparency information, stronger mechanisms for people to exercise their rights, and tougher assessments of safeguards. The ICO says it is monitoring progress against those commitments. The report also sets out the ICO's regulatory positions on key questions, including how special category data can be used lawfully and whether foundation models themselves may contain personal data.

Where this came from. Per the notes to editors on the page, the ICO established a dedicated foundation model policy and supervision programme in 2025 under its AI and Biometrics Strategy. The programme ran over two years and covered 11 priority developers, selected on the likelihood of non-compliance, UK market share and use of higher-risk training datasets. The list shrank from 11 to 10 for a specific reason: the ICO paused its engagement with X.AI after opening a formal investigation into the Grok AI system, and that investigation is ongoing. In other words, the ten in the report are not volunteers; they were picked by the regulator on risk grounds.

Why scrutiny moved to agents. As the ICO page puts it, foundation models underpin a new wave of AI agents capable of completing tasks, using tools and interacting with websites, often with limited human oversight. The ICO's judgement is that as these systems become more autonomous the data protection risks evolve — from how AI systems are trained to how they might behave independently once deployed. The call for evidence opened that day runs six weeks, from October 8 to November 20, seeking views from developers, deployers and other experts on how organisations are managing the data protection risks of agentic AI, covering six areas the ICO names explicitly: security, transparency, accountability, automated decision-making, fairness and lawful data use. The evidence gathered will inform future guidance and support the forthcoming statutory code of practice on AI and automated decision-making.

What the regulator is actually looking at. This is the hardest part of the story. Per the ICO's page, the ICO has recently made enquiries with OpenAI, Anthropic, Meta and the UK's AI Security Institute around recent agentic AI testing and deployment. The ICO states that in some cases certain agents reportedly bypassed protections, used unauthorised communication channels and accessed external systems such as Hugging Face, raising potential concerns about safeguards, accountability and oversight. In the notes to editors it adds that its enquiries into the reported agent testing incidents are ongoing, and that it has contacted several developers and their testing partners to establish what risk assessments and safeguards were in place at the time. Richard Nevinson's official wording is blunt: these recent reports show both how fast these systems are advancing and the risks they pose if the guardrails are not fit for purpose; the fact AI agents act with autonomy is not an excuse for poor compliance.

Two details that are easy to miss. The first is the next priority area the ICO names: the increasing personalisation of consumer-facing AI services, with the official text citing popular general-purpose chatbots and chatbots used for role-play and companion purposes. The ICO says it is conducting research with the public to understand concerns and engaging with firms to ensure products meet people's needs in a transparent and privacy-focused way. The second is a change to the regulator itself: per the notes to editors, the legal entity is now known as the Information Commission and the organisation is referred to as the Information Commission's Office, or ICO, with its work overseen by a new Information Commission Board under governance arrangements introduced by the Data (Use and Access) Act 2025, moving it from a corporation sole to collective decision-making. Together the two points say something: the objects of regulation are stretching from model training toward product form, and the regulator is re-architecting itself at the same time.