
Key takeaways
- The Commission presented the Digital Fairness Act on 11 November 2026. It still has to pass the European Parliament and member states, which dpa says could take months or years.
- Reports on the proposal say customers would always be able to speak to a human about a problem with a contract, and AI-only customer service lines would be banned.
- People would keep their consumer rights when an AI agent buys on their behalf, and should stay in control, for example by setting price limits.
- If you sell to EU customers, the practical move now is to make sure your chatbot has a clear route to a person and your checkout has no pressure tricks.
What the Commission proposed
Commissioner Michael McGrath and Executive Vice-President Henna Virkkunen presented the Digital Fairness Act in Brussels on Wednesday, 11 November. It’s a broad consumer-protection package. Eunews reports it covers influencer marketing, targeted ads aimed at minors, personalised pricing, dark patterns, addictive design and subscriptions that are hard to cancel.
Two parts matter most for anyone running AI in sales or support. One is customer service. The other is AI agents that shop for people. We cover both below.
Enforcement gets sharper too. The Commission wants its own powers to investigate large cross-border cases, and fines could reach 6% of a company’s worldwide annual turnover. Eunews also reports that companies outside the EU would have to appoint a legal representative inside the Union, so US and Gulf businesses selling into Europe aren’t exempt.
A right to reach a human
According to dpa’s report, anyone who has concluded a contract should always be able to speak to a human about problems, and the Commission wants to ban customer service hotlines run only by technology such as AI. Eunews describes the same principle: consumers must be able to reach a human representative.
This doesn’t ban chatbots. It bans chatbots with no way out. A bot that answers order-tracking questions at 2am and then hands a frustrated customer to a person is fine under this logic. A bot that loops people round the same three answers isn’t.
We’ve seen the same idea elsewhere. New York City has proposed rules aimed at chatbot “doom loops”, which we covered in our article on the NYC proposal. Regulators on both sides of the Atlantic are landing in the same place.
AI agents that shop for customers
The second change looks further ahead. dpa reports that consumers would keep all their rights when an AI agent shops online for them, and that they should stay in control of what the agent does, for example by setting price limits.
That lines up with what big retailers are already building. Walmart, for instance, has announced agent profiles with spending limits or approval for every order. If you run an online store, expect agent-placed orders to carry the same refund, cancellation and information rights as any other order.
There’s also an earlier detail worth knowing. Before the official launch, Bloomberg reported (via AI Weekly) that a leaked draft named “AI chatbot functions that simulate interpersonal relationships to prolong engagement” as a feature adults must be able to switch off. The final reports we’ve read don’t mention chatbots by name, so treat that as unconfirmed until the full legal text is checked.
What to do now
Nothing here applies yet, and the text will change in negotiation. But most of these steps are good practice anyway, so they won’t be wasted:
- Map every way a customer can contact you and check each one has a route to a person within a reasonable time. Our guide to chatbot human handoff walks through the design.
- Audit your checkout and sign-up flow for fake scarcity, pre-ticked add-ons and trials that turn into paid plans without a clear prompt.
- Make cancelling as easy as subscribing. If it takes one click to join, it shouldn’t take a phone call to leave.
- Decide how you’ll treat orders placed by AI agents: what you’ll accept, how you’ll verify them and how refunds work.
- Keep an eye on the AI Act too. The Digital Fairness Act will sit alongside it; our Digital Omnibus explainer covers the AI Act timeline.
If you’re building or replacing a support bot, it’s far cheaper to design the human handoff in from day one than to bolt it on later. Our AI chatbot development work starts with exactly that question, and our UK and Europe consultancy page explains how we work with EU-facing businesses.
This article is general information, not legal advice. Check the final legal text, and speak to a qualified lawyer about your own situation.
Frequently asked questions
Is the Digital Fairness Act law yet?
Does it ban AI chatbots in customer service?
Does it apply to businesses outside the EU?
What are the fines?
Sources
- Eunews: EU unveils Digital Fairness Act (11 Nov 2026)
- dpa via Yahoo News: EU executive unveils raft of measures to protect consumers online (11 Nov 2026)
- AI Weekly, citing Bloomberg: EU draft Digital Fairness Act targets AI companion bots (9 Nov 2026, based on a leaked draft)