Regulatory  ·  2026-11-21

UK regulators (CMA, FCA, ICO, Ofcom) issue joint DRCF letter setting expectations for personal AI agents; existing law applies in full

RegulatoryHigh impactUnited Kingdom
On 16 November 2026, the heads of the CMA, FCA, ICO and Ofcom (via the Digital Regulation Cooperation Forum) sent a joint letter to the Chancellor, the DCMS Secretary, the BIST Secretary and the Minister for AI. It warns that several personal AI agent products are about to launch in the UK, that firms must meet legal and regulatory obligations and put safeguards in place before launch, and that engagement 'has not been consistently early or open across the relevant regulators.' It identifies fraud/scams, unexpected personal-information sharing and misleading recommendations as key risks, requires that agents give users agency and not compromise operational resilience, and states regulators 'will not hesitate to use our powers' where law is breached. It builds on the 30 October joint statement that an agent's autonomy does not reduce organizational responsibility, and confirms financial-services and telecoms firms' own obligations continue to apply.
This is the clearest cross-regulatory statement yet on the agentic economy in the UK: consumer, competition, data-protection, online-safety, financial-services and payments law all apply to AI agents, with active enforcement promised. It imposes an early-engagement, impact-assessment and post-launch-testing discipline on any firm bringing an agentic product to market (or integrating third-party agents) in the UK, and effectively sets the governance bar in a jurisdiction without a standalone AI statute.
Firms building or deploying personal AI agents in the UK should identify the lead regulator(s), document what the agent accesses/shares/acts on, prepare impact assessments and safeguards, treat model/feature updates like a fresh launch, and engage the relevant regulator before launch — especially financial-services firms partnering with agent providers.
DRCF — joint letter on personal AI agentsAndres Lehtmets analysisCaptain Compliance
See this in the live feed Explore related AI security and governance findings — updated every morning.
Open the feed →