Should your AI say it’s an AI? Disclosure rules for sales conversations
Since August 2026, AI systems that talk to people in the EU have to say they’re AI. What that means for UK SaaS teams, and why being upfront is good sales practice anyway.
What you will learn
- What does the EU AI Act require when an AI talks to a buyer?
- Do the rules apply to UK companies selling into Europe?
- Does telling buyers they’re talking to AI hurt conversion?
In this article
AI that can hold a natural sales conversation raises an awkward question. If a buyer can’t tell they’re talking to software, should you tell them?
In Europe, that’s no longer a judgement call.
What the EU AI Act requires
Article 50 of the EU AI Act sets out transparency rules, and the duty to tell people they are interacting with an AI system has applied since 2 August 2026. The EU’s “digital omnibus” package, agreed in May 2026, postponed several other parts of the Act but left this disclosure duty on its original timetable.
In practice, the key points are:
- Tell people at the first interaction. The disclosure has to be clear and hard to miss. A line buried in your terms doesn’t count.
- The exception is narrow. You can skip it only where it’s obvious to a reasonable person that they’re dealing with AI. A natural-sounding voice agent is precisely the case where it isn’t obvious.
- Penalties are significant. Breaches can attract fines of up to €15 million or 3% of worldwide turnover.
Does this apply to UK companies?
Often, yes. The Act reaches providers outside the EU where their AI’s output is used in the Union. If your UK SaaS business takes demo requests from buyers in Dublin, Amsterdam or Berlin, you should assume the rules apply to those conversations.
For purely UK buyers, there’s no direct equivalent law yet. But the UK’s general rules against misleading people still apply, and data protection law already expects you to be open about how you process people’s information. Pretending an AI is a person is hard to square with either.
We’re not lawyers, so take specific advice for your situation. The direction of travel, though, is clear.
Disclosure isn’t the risk. Getting caught pretending is.
Won’t disclosure put buyers off?
It’s a fair worry, but the evidence points the other way. Buyers are already sceptical: in Gartner’s 2026 research, about half of B2B buyers said they expect misleading information from AI tools, and almost as many said the same of salespeople.
Trust is the scarce resource. A buyer who discovers halfway through a call that they’ve been talking to software feels tricked, and that feeling attaches to your brand. A buyer told upfront can decide for themselves, and most will judge the conversation on whether it was useful.
What good disclosure sounds like
- Say it first. Name the AI as an AI in the opening line, before any questions.
- Keep it natural. “I’m [Company]’s AI assistant. I can show you the product and answer your questions now” works better than a legal notice.
- Offer a human. Make it easy to book a person, and mean it.
- Be honest about limits. If the AI doesn’t know something, it should say so and pass the question on.
- Remind people where it matters. In longer conversations, or when pricing or contracts come up, a quick reminder keeps things clear.
The bottom line
The law now requires disclosure for many sales conversations, and good selling required it already. Tell buyers they’re talking to AI, make the conversation worth having, and hand over to a person when it counts.