From August 2, Chatbots in Poland Must Disclose They Aren't Human

From August 2, 2026, every company operating in Poland and the European Union that serves customers through a chatbot, voicebot, or virtual advisor will have to clearly tell the person on the other end that they're not talking to a human. It's the next stage in rolling out the EU's AI Act, this time targeting not the makers of the most powerful models, but ordinary businesses using off-the-shelf AI tools in customer contact.

This isn't a last-minute addition. Article 50 of the AI Act has long been on the EU's implementation schedule, but only now, with the deadline actually approaching, are companies starting to tally the cost of adapting their systems. As lawyers cited by Polish trade media point out, many businesses still don't realize they're even using solutions covered by the regulation.

What Companies Must Do

The requirement sounds simple, but in practice it means changes in many places at once. Providers of systems meant for direct contact with people must design them so users know immediately they're talking to artificial intelligence. The form of the notice can vary by channel: a short line of text is enough in a chat window, a phone hotline will need a spoken announcement, and a voice conversation needs a separate audio cue to distinguish the bot from a human.

The disclosure can't be buried in a multi-page terms-of-service document or revealed only after the conversation ends. It must be clear, recognizable, and accessible, including to people with disabilities. That sets the rule apart from the practice many companies have followed so far, mentioning AI in their terms while saying nothing about it during the actual interaction.

Who's Actually Affected

The scope is broader than it might first appear. This isn't just about chatbots on online store websites. The rule also covers automated phone agents in call centers, virtual advisors in banking apps, bots that walk customers through complaint procedures, and systems that respond on behalf of public administration offices.

That means practically every medium and large company in Poland that has automated its customer service in recent years now has to check whether its setup meets the new transparency standard. Banks, insurers, telecom operators, and online retailers are among the groups most exposed to scrutiny, since those industries have pushed customer-contact automation the furthest.

Different From Penalties for Model Makers

It's worth distinguishing this obligation from the European Commission's widely publicized plan to penalize makers of the most powerful AI models, which also takes effect in August. That mechanism targets providers of general-purpose models that pose systemic risk, in practice a handful of the world's largest tech companies. Article 50 works differently: it applies to any entity deploying a ready-made conversational tool, regardless of size or whose AI model powers the chatbot.

For a Polish business running an online store or service company, that means the responsibility falls on them directly, not on the technology provider. Using an off-the-shelf chatbot bought from an outside vendor doesn't exempt a company from the obligation to clearly disclose that customers are talking to a machine.

What About Other Regulations

AI Act rules don't replace other regulations. Personal data protection, consumer rights, banking secrecy, and medical confidentiality all apply in parallel and independently of the new transparency requirement. Simply telling users they're talking to AI doesn't automatically legalize every practice around processing their data, nor does it create default consent for automated data processing.

That's an important caveat, since some companies may have hoped that adding a notice about AI would resolve every legal question raised by deploying a chatbot. In practice, it means double the work: companies need to comply with the new AI-disclosure requirement while also maintaining compliance with existing data protection rules.

A conversation with AI must be clearly labeled no later than the moment of first contact, and the disclosure can't be hidden in the terms of service or revealed only after the conversation ends - Forsal.pl, based on its analysis of Article 50 of the AI Act

What Comes Next for Companies

There's little time left before August 2, and according to the legal analyses cited, a significant share of Polish companies still haven't audited their systems for compliance with the new requirement. Experts recommend starting with a simple review: identify where automated customer-contact systems operate within the company, and add a clear notice at every one of them that the party on the other end is artificial intelligence. The cost of making the change itself is often small, but the cost of not making it, given penalties reaching into the tens of millions of euros, could be severe even for a mid-sized company.

Sources: Are You Talking to a Human or a Machine? Starting August 2, Companies Must Clarify (forsal.pl), Starting in August, the EU Ends the Pretense That AI Is Human (dziennik.pl)