"Am I talking to a real person?" is a question customers are asking more often, and it now sits closer to law than manners. The ai disclosure rules uk business owners need to understand come from a real, signed statutory instrument — not a guideline someone might eventually enforce. If your business uses an AI voice agent, an AI chatbot, or AI-generated content anywhere a customer might reasonably be misled, this is worth fifteen minutes of your attention now rather than a difficult conversation with a regulator later.

The short answer: yes, and increasingly it's the law, not just good practice

If an AI system is handling any part of a customer interaction — answering the phone, chatting on your website, drafting the words in your adverts — the safe, compliant and genuinely low-risk position is to say so clearly. Two pieces of UK law, both now in force, point in the same direction, and the advertising regulator has already made its own position clear on top of them.

SI 2026/425: the regulation actually driving this

Most of the specific ai disclosure rules uk business owners are now hearing about trace back to a single piece of legislation. The Data Protection Act 2018 (Code of Practice on AI and Automated Decision-Making) Regulations 2026 — known as SI 2026/425 — was signed on 16 April 2026, laid before Parliament on 21 April, and came into force on 12 May 2026. You can read the instrument itself on legislation.gov.uk.

What it actually does is require the Information Commissioner's Office to produce a statutory code of practice covering AI and automated decision-making. That's a real, binding legal duty on the regulator — it is not the same as the finished code already existing. The ICO's consultation on the draft closed on 29 May 2026, and the full code text is expected later in 2026 or in 2027. Businesses building or buying AI voice systems now should treat the direction of travel as settled even while the fine detail is still being finalised.

The ICO's own hub on AI and data protection guidance is the place to check for updates as the code develops — it's a live page, not a one-off announcement, and it's worth bookmarking rather than relying on a single article to stay current.

What the direction of the code points towards for voice AI

Based on the consultation direction and existing regulatory expectations for automated decision-making, businesses using AI to handle customer calls should reasonably expect obligations along these lines once the code is finalised:

Audible disclosure, early in the call

A caller should be told they're speaking with an AI system near the start of the interaction — not buried in a terms page they'll never read, and not left to guess from an unnatural pause.

A record of what was decided and why

Where an automated system makes any kind of decision affecting a customer, there should be a way to explain, after the fact, what happened and on what basis.

Fair treatment regardless of accent or background

An AI voice system that performs noticeably worse for some accents or demographics than others is a real bias problem, not a technical footnote — expect scrutiny here to increase, not soften.

A working route to a human

Disclosure on its own isn't the whole job. A customer who knows they're speaking to an AI still needs a genuine, unblocked way to reach a person when the situation calls for it.

The Data (Use and Access) Act 2026 — the other half of the picture

Alongside SI 2026/425, the Data (Use and Access) Act 2026 came into force earlier in the year, on 5 February 2026, and reshapes some of the UK's automated decision-making safeguards that previously sat under Article 22 of the UK GDPR. The practical effect for most small businesses is the same as SI 2026/425 points towards: if a decision affecting a customer is made automatically, there needs to be transparency about it and a route to human review. The government's plain-English guidance on data protection and your business is a reasonable starting point if this is the first time your business has had to think about automated decision-making at all.

It isn't only about phone calls

The Advertising Standards Authority has already set out its own test for AI in marketing, separate from the data protection rules above: would the audience be misled if the use of AI weren't disclosed? The ASA has been clear that disclosure cannot rescue fundamentally deceptive messaging, and that advertisers remain responsible for what their AI tools produce — an AI-written ad claim that's misleading is still the advertiser's problem, not the tool's. If your business uses AI to draft ad copy, website content or social posts, this sits alongside the voice-specific rules above as a second, related exposure worth checking.

A practical disclosure checklist for AI phone answering

Turning the direction of the law into a working checklist is the fastest way to act on ai disclosure rules uk business owners can apply today, without waiting for the finished code.

Requirement	What it looks like in practice
Say it early	State clearly, in the first sentence, that the caller is speaking with an AI assistant
Keep it honest, not hidden	No vague phrasing designed to leave the caller unsure
Give a real route to a person	A clear, working way to reach a human, not a dead end
Limit it to approved information	The system only answers using information the business has actually approved
Keep a record	Be able to show what the system said and why, if ever asked

How this actually sounds on a real call

This isn't theoretical for us — it's how our own AI voice agent is configured for every First Essential UK client. The very first thing it says is some version of "you're speaking with an AI assistant." No trick, no hiding it, no hoping the caller doesn't notice. It handles routine questions — hours, availability, simple bookings — using only information the business has approved, and the moment something needs judgement, it hands over to a person immediately. See our guide on how an AI voice agent actually works for the full breakdown of that handover process.

Why disclosure is a sales asset, not a weakness

The instinct for a lot of business owners is to worry that telling a customer "this is AI" will put them off. In practice, the opposite tends to be true. Callers who know upfront what they're dealing with trust the rest of the call more, not less — and a business that's upfront about its own AI use looks like it has nothing to hide, which is a genuinely useful thing to project in a market where AI missteps make the news regularly.

There's also a real competitive gap here. Almost nobody selling AI phone answering below roughly £299 a month builds a proper disclosure script, an escalation policy and a basic record of automated decisions into what they sell. Treating compliance as a standard deliverable rather than an afterthought turns your biggest objection — "is this even allowed?" — into the reason a cautious customer chooses you over a cheaper competitor.

What happens if you get this wrong

Getting AI disclosure wrong isn't primarily a fines risk for most small businesses at this stage — it's a trust and complaints risk. A customer who feels deceived is more likely to complain, more likely to leave a negative review, and less likely to give the business a second chance. The Citizens Advice guidance on getting help with a consumer issue is a useful reminder of the kind of route a dissatisfied customer already has open to them, disclosure rules aside.

Escalation: the part disclosure alone doesn't solve

Telling a customer they're speaking with an AI is only half the job if there's no genuine way for them to reach a person when they need one. See our guide on how AI call handling should hand over to a human for what a working escalation route actually needs to include — it's the practical companion to everything above, not a separate topic.

Fair treatment across accents and demographics

An AI voice system that consistently mishears certain accents, or handles some customer groups worse than others, is a fairness problem that sits close to existing equality law, not just an emerging AI-specific rule. The Equality Act 2010 and its accompanying official guidance already set expectations about not disadvantaging customers based on protected characteristics — an AI system that performs unevenly by accent can fall foul of the same principle, regulatory code or not. Testing your own system against a genuinely diverse range of real callers, not just the accents easiest to test with, is worth doing before problems surface in a complaint.

Building disclosure into every customer-facing AI use, not just phone calls

The same logic applies beyond voice. A website chatbot, an unattended kiosk, and outbound calling all raise the same basic question — does the customer know they're dealing with AI, and can they reach a person if they need to? Our guide on AI making outbound calls to your customers covers the specific disclosure considerations that apply when your business is calling out rather than answering in.

If you're weighing up an AI answering service against a traditional one, our comparison of answering services versus AI covers where disclosure fits into that broader decision, alongside cost and coverage.

Getting accuracy right alongside disclosure

Disclosure doesn't fix a system that regularly mishears customers or hands over clumsily — accuracy and honest handover need to work together. See what actually affects AI voice agent accuracy for the practical side of getting the technology itself right, not just the compliance wrapper around it.

For a wider view of where AI genuinely helps a business owner day to day — and where it deliberately doesn't — our guide on what an AI assistant actually does for a business owner sets out the same "approved information only, human handover when needed" principle that underpins everything in this article.

Data protection basics that sit underneath all of this

Every AI phone system also processes personal data — the caller's voice, whatever they say, and any details captured. The ICO's general UK GDPR guidance for organisations remains the baseline requirement regardless of the AI-specific code still being finalised, and basic account and system security matters here too — the National Cyber Security Centre's small business guide is a sensible starting point for a team without a dedicated IT function.

What to do this week, before the full code lands

Acting on ai disclosure rules uk business owners will eventually be assessed against doesn't require waiting for the ICO's finished code. Write a one-line disclosure script and make sure your AI system actually says it, first sentence, every call. Confirm there's a genuine, tested route to a human. Keep a basic note of what your AI system is and isn't allowed to say. None of this requires new software — it requires a decision, written down, and checked.

Key takeaways

• SI 2026/425 is a real, signed regulation in force since 12 May 2026 — it requires the ICO to produce a statutory code on AI and automated decision-making, with the full code still being finalised.
• The direction is clear even before the detail lands: audible disclosure, a record of automated decisions, fair treatment across accents, and a genuine route to a human.
• The Data (Use and Access) Act 2026 (in force 5 February 2026) reinforces the same transparency principle for automated decisions generally, not just voice calls.
• The ASA's separate test for AI in advertising is simple: would the audience be misled if AI use weren't disclosed?
• Disclosure is a trust-builder and a sales asset, not a weakness — almost nobody in the budget tier of AI phone answering builds it in properly.

Frequently asked questions about ai disclosure rules uk business owners face

Is AI disclosure to customers already a legal requirement in the UK?

SI 2026/425 is in force and requires the ICO to produce a statutory code covering AI and automated decision-making. The full code's detailed requirements are still being finalised, but the current direction of the ai disclosure rules uk business owners will face — audible disclosure, fair treatment, a human escalation route — is already clear enough to act on now.

Does this only apply to AI phone systems?

No. The same underlying principles apply to chatbots, self-service kiosks, outbound AI calling, and AI-generated marketing content, which the ASA already regulates separately under its own disclosure test.

What should an AI voice agent actually say to disclose itself?

A short, clear statement near the start of the call — something like confirming the caller is speaking with an AI assistant — is the safe, low-risk approach, paired with a genuine route to a person.

Could failing to disclose AI use lose me customers rather than just risk a fine?

For most small businesses, the more immediate risk is trust and complaints, not enforcement. A customer who feels misled is more likely to leave a negative review than to report you to a regulator.

Do I need to wait for the ICO's final code before doing anything?

No. Writing a disclosure script, testing your escalation route and keeping a basic record of what your AI system is allowed to say are all sensible now, regardless of when the finished code lands.

Is disclosure bad for conversion or customer experience?

The evidence so far points the other way — callers who know upfront what they're dealing with tend to trust the rest of the interaction more, not less.

Ready to check your own AI disclosure setup?

We build audible disclosure and a genuine human handover into every AI voice agent we configure. Talk through your setup or see our AI Solutions.