This browser is not actively supported anymore. For the best passle experience, we strongly recommend you upgrade your browser.
| 3 minute read

Autonomy (but not immunity) for AI agents – regulators expect early engagement

The UK’s Digital Regulation Cooperation Forum (DRCF), which brings together the CMA, FCA, ICO and Ofcom, has made clear that businesses developing or deploying personal AI agents are expected to engage early with regulators to ensure they meet their legal and regulatory obligations.

In a letter to ministers on 16 November, the DRCF recognised the transformative potential of personal AI agents, while underscoring that firms considering deploying personal AI agents in the UK (including those partnering with AI agent providers) should engage “early and openly with regulators before bringing their products to market”. The regulators aim to ensure that businesses have factored in compliance with existing laws, including on competition, consumer, data protection, online safety, financial services, payments and cyber security, along with “any necessary authorisation or registration”. The statements were reinforced by a speech by the CMA’s CEO on 17 November.

The DRCF regulators have backed this open approach with a warning that they “will not hesitate” to use their sanctioning powers where companies are reasonably suspected to have breached these rules, risking harm to consumers or businesses. 

What has prompted this? 

The letter cites the expected launch of numerous personal agentic products in the UK – a “step change” in the way AI is being used – as the trigger for the regulators’ message. The letter, together with an earlier statement (published 30 October), sets out the DRCF’s expectation of early and open engagement, coupled with a warning of regulatory action where conduct breaches the law. Earlier in November, the DRCF also published its Consumer Interest and AI report exploring consumer attitudes to the risks and benefits of using AI.

This cross-regulatory approach reflects regulators’ concerns that the rapid growth and use of agentic products could introduce new risks and amplify existing ones for consumers. The DRCF seeks to balance encouraging “responsible innovation” with ensuring that businesses understand their obligations and put appropriate safeguards in place to mitigate risk.

The DRCF regulators are also keen to improve their “understanding of emerging technologies”. Through an “open door” approach, they aim to keep pace and stay informed on what businesses are doing, allowing appropriate testing of new products before they are deployed. 

What are the regulators’ concerns? 

The DRCF recognises that AI agents could deliver significant benefits for consumers including cost and time savings, accessibility and personalised advice or recommendations. But it also points to the substantial risks that arise if agentic AI is not implemented responsibly. 

Some of the concerns highlighted by the DRCF include:

  • how agents act as intermediaries for customers, including:

    • how AI agents operate with commercial partners

    • whether smaller suppliers can compete for visibility 

    • how incentives are disclosed to customers

  • the opacity of agentic systems (making it difficult for customers to fully exercise control), 

  • risk around fraud, scams and manipulation through misleading information,

  • risk of inaccuracies or hallucinations,

  • the operational resilience of the systems which AI agents may interact with,

  • risks of agentic collusion, and

  • data security and privacy.

What are regulators currently doing about their concerns? 

In their joint statement, DRCF regulators make clear they are tracking market developments closely and engaging directly with companies who are developing and deploying AI agents, using regulatory sandboxes to test products. Alongside individual publications by the regulators (such as CMA research and guidance on agentic AI and FCA good and poor practice for AI adoption), through the cross-regulator forum they have been sharing intelligence and evidence, with close coordination at leadership level. 

This follows developments at an EU level, where recent proposals on the Digital Fairness Act have been strengthened to include obligations regulating the use of AI agents in e-commerce. These include (i) providing consumers with increased control over those agents, and (ii) providing transparency in relation to contracts formed using the system.

What should businesses do?

If you are considering developing or deploying a personal AI agent, early engagement with the relevant regulator is important, taking advantage of their open-door approach. 

Preparation will be key, including in the review of recommendation design and commercial incentives. For example:

  • Is there a clear target market and intended outcome for the agentic product?

  • Does the agent search widely, or favour its provider’s partners? Do commercial arrangements constrain consumer choice?

  • Is there transparency for consumers as regards the role, commercial interests and risks of the agent (including any regulatory protections, or absence thereof)?

  • How is personal data used?

  • Who is accountable for the agent’s decisions?

  • What governance has been built around the agentic AI to oversee the effectiveness of compliance controls?

  • What is the process when something goes wrong?

Businesses should ensure appropriate safeguards and controls are put in place, which could include:

  • spending limits,

  • approval for consequential actions,

  • accessible stop controls,

  • clear internal ownership of the agent,

  • audit trails or decision logs, and 

  • a clear complaints route for the public.

To stay up to date with the latest tech developments - subscribe now!

Tags

ai, antitrust & foreign investment, consumer protection, data and cyber