The EU AI Act introduces a transparency rule from 2 August 2026: customers must be told when they're interacting with AI. It applies automatically if you serve EU customers, and is recommended best practice for UK-only businesses. Standard automation, rule-based workflows and call transcription aren't affected. Only bots that talk to customers directly need an updated greeting or footer.
✓ If you serve EU customers, the new rules apply to you from 2 August 2026, wherever your business is based.
✓ UK-only businesses aren't legally required to act, but it's recommended good practice under consumer protection and GDPR principles.
✓ The core rule in one line: customers need to know when they're talking to AI, not a person.
✓ Not everything automated counts as AI. Rule-based workflows, automatic replies, caller ID and standard reporting are unaffected.
✓ Voicebots, chatbots and emailbots need a short, plain AI disclosure at the start of the interaction.
✓ Background AI transcription isn't covered by the new disclosure rule. Your existing GDPR call recording notice still applies.
✓ Daktela's AI tools are already built for this: UK/EU-hosted, data never used to train public models, fully confidential.
First things first: does this even apply to you?
If you serve customers in the EU, yes, the new rules apply to you from 2 August 2026, regardless of where your business is based. The law follows the customer, not the company.
If you only serve UK customers, there's no standalone UK law requiring this yet, so you're not under a legal obligation in the same way. That said, we'd still recommend treating it as good practice under UK consumer protection and GDPR transparency principles, it costs you a sentence and it heads off any risk of looking like you're hiding something from customers.
Either way, here's what's actually changing and what isn't...
It's not a crackdown on AI. It's a transparency rule.
The whole thing boils down to one sentence: customers need to know when they're talking to AI, not a person. That's it. It's not about banning automation or slowing down innovation, it's about being upfront.
What stays exactly the same?
Not everything automated counts as "AI" under the rules. Rule-based workflows, automatic replies, caller ID routing, keyword-triggered responses, and your standard reporting all carry on working exactly as they do now. No new obligations here.
What needs a small update
If you're using voicebots, chatbots, or emailbots to interact directly with customers, these are the systems the new rules are aimed at. For anyone in scope, customers need to be told clearly, at the start of the interaction, that they're dealing with AI.
That doesn't mean a wall of legal text. A short, plain statement does the job.
For example:
Simple, honest, and it builds trust rather than eroding it.
What about call transcription?
Good news here: standard AI-powered speech-to-text, transcribing calls in the background, isn't treated the same way as a bot that talks to a customer directly. It's generally understood to sit outside the new disclosure requirement, since it's a background process rather than something the customer interacts with.
What still applies is what's always applied: your standard call recording and transcription notice under GDPR. If you're already telling customers that "calls may be recorded and transcribed for quality and training purposes," that continues to do the job.
What doesn't change: your data
Daktela's AI tools are already built for this. Your data is processed and stored in secure, certified data centres in the UK and EU, it's never used to train external or public AI models, and it stays confidential to your organisation. Nothing about the new rules changes how we handle it.
A quick checklist
Bottom line
This isn't the end of automation, it's a framework for using it responsibly. A few updated greeting scripts and footers, and you're covered.
This article is general guidance, not legal advice. Please check your specific setup with your own legal counsel.
