AI Agents and the ICO: What the New Call for Evidence Means for Your Firm
United Kingdom: the AI compliance framework that answers this
£797/year
The ICO has opened a call for evidence on AI agents, closing 20 November 2026, and says autonomy is no excuse for poor compliance. What professional services firms should do now.
The UK's data protection regulator has started work on how the law applies to AI agents. These are AI tools that don't just answer questions but go and do things: they send the email, book the meeting, update the record or chase the invoice, often with nobody checking each step.
On 8 October 2026 the ICO opened a six-week call for evidence on agentic AI. It closes on 20 November 2026. The regulator wants to hear from the people who build these tools and also from the businesses that use them.
What the ICO is asking about
The call for evidence covers six areas:
- Data security: how you stop an agent exposing or leaking personal data.
- Transparency: whether people know an agent is handling their information.
- Accountability: who in your business answers for what the agent does.
- Automated decision-making: when an agent makes a decision about a person with no human involved.
- Fairness and purpose limitation: whether the agent uses data only for the reason it was collected.
- Lawfulness of processing: whether you have a lawful basis for what the agent does with personal data.
The ICO says the answers will shape its future guidance and its forthcoming statutory code of practice on AI and automated decision-making. A statutory code carries weight: it is the regulator's own account of what compliance looks like, and it is what you would be measured against.
The ICO has already given its starting position
On the day it opened the call, the ICO said that "the fact AI agents act with autonomy is not an excuse for poor compliance". The regulator also reported that in some testing, agents had bypassed the protections built around them.
For a firm, the message is simple. If an AI tool acts on your behalf, the responsibility is still yours. "The software did it" is not a defence.
Why this matters for professional services firms
Accountants, solicitors, agencies, recruiters and HR consultancies are taking on agent-style tools quickly. Inbox assistants draft and send replies. Copilot-style agents work across client files. Recruitment platforms screen and rank candidates. Most of these handle client or candidate personal data, and some make or shape decisions about people.
None of this is banned. But the regulator is now looking closely at exactly how firms keep control of it.
What to do now
- List the AI tools that act on their own. Leave out the ones that only answer a question you typed. Include the ones that send, book, file, score or decide.
- Name an owner for each one. Pick one person who can say what the tool is allowed to do, and stop it if needed.
- Write down what personal data each tool touches. Note where that data goes, including any supplier outside the UK.
- Check for decisions about people. If a tool decides something significant about a client, candidate or employee without a person involved, you need safeguards. Those include letting the person ask for a human to look at it again.
- Assess the high-risk uses. UK GDPR already requires a data protection impact assessment where processing is likely to be high risk, and an autonomous tool working through client files can be exactly that.
- Consider responding. The call is open to businesses that use these tools, not only the developers. If your firm has a view on what workable guidance looks like, 20 November is the deadline.
A note on the regulator's name
Since 30 September 2026 the legal body has been the Information Commission, run by a board, and no longer a single Commissioner. It is still known as the ICO, and its powers over your business have not changed.
Where we come in
Our UK AI compliance framework turns most of this list into documents your firm can actually use. It includes a register of who and what is allowed near your data, the points where a person must check an AI tool's work, and a self-assessment against the ICO's AI guidance. The Complete tier adds a risk register for your current tools, plus notices and opt-out procedures for automated decisions. Legal updates are included for the year, so when the ICO publishes its guidance and code, your framework moves with it. See the UK framework.
Follow us in Google
See Ops Intel first when AI rules change
One click tells Google you want our compliance briefings near the top of your results — and marks us as a preferred source inside Google's AI answers. It applies to your Google account only, and you can undo it any time.
What to do about it
The news is what changed. A framework is what you do about it.
Ops Intel writes AI compliance frameworks for small and medium businesses worldwide. Before you spend anything, read a real one — the whole pack, produced by the same system that will write yours.