EU AI Act Forces Polish Companies to Disclose AI Chatbots From August 2

Starting August 2, 2026, companies operating in Poland and across the European Union will no longer be able to let customers guess whether they are talking to a human or a machine. Article 50 of the EU's Artificial Intelligence Act, known as the AI Act, requires businesses to clearly and accessibly inform users that they are interacting with an AI system, at the latest by the time of the first interaction.

What exactly is changing

Until now, many companies buried in their terms and conditions the fact that an algorithm, not a live agent, was responding on the other end of a chat or hotline. After August 2, such practices will be unlawful. The notice that a customer is dealing with AI must appear directly alongside the conversation itself, for example as a short message above the chat window, a voice announcement before a phone call begins, or a clear notification before the interaction starts.

The requirement covers more than classic chat windows on websites. It also applies to automated phone consultants on hotlines, virtual advisors in banking apps, bots guiding customers through complaint procedures, and systems responding on behalf of government offices. Practically every company in the financial, e-commerce, or public services sector that has deployed AI-based customer service needs to review its interfaces before August.

Accessibility and exceptions

The rule also requires that notice of talking to a machine be accessible to people with disabilities, through appropriate contrast, readable text, and audio versions suited to the communication channel. Fine print in a website footer will not suffice if it fails to meet digital accessibility standards.

Lawmakers also carved out exceptions. The obligation does not fully apply to systems used by law enforcement authorities to detect and investigate crimes, provided appropriate safeguards for individual rights are maintained. The situation is also different when artificial intelligence merely supports an employee in the background without presenting itself to the customer as an independent conversational partner. In that case, the obligation under Article 50(1) does not have to apply in the same form, although other data protection obligations remain in force.

The cost of non-compliance

The financial stakes are high. Violating the transparency obligations of Article 50 can cost a company up to 15 million euros or up to 3 percent of its total global annual turnover from the previous financial year, whichever is higher. Smaller companies are subject to the lower of these two values. When setting the fine, regulators will consider the severity and duration of the violation, the number of people affected, and any corrective action taken by the company.

What Polish companies need to do

By August 2, companies should review all of their customer touchpoints and determine where AI operates independently and where it merely supports an employee. Interfaces and call scripts need to be updated so the AI notice appears in the right place, and escalation procedures to a human need to be put in place for when a customer requests one. For many mid-sized companies, banks, insurers, and online stores, this amounts to a real implementation project, not just a cosmetic change to the terms of service.

The deadline coincides with a growing wave of AI agent deployments at Polish customer service operations, from bank hotlines to e-commerce complaint bots. New-technology law practitioners expect a wave of compliance audits in July, as the Urząd Ochrony Danych Osobowych (Poland's data protection authority) and other regulators have signaled they will treat this obligation as a priority from day one.

Sources: AI Act a chatboty. Obowiązek informowania użytkowników od 2 sierpnia 2026 r. (forsal.pl), Od sierpnia w UE kończy się udawanie, że AI to człowiek (technologia.dziennik.pl)