Starting August 2, EU Chatbots Must Disclose They Are AI

Starting August 2, 2026, every company operating in the European Union that uses chatbots, voicebots, or automated consultants will have to clearly tell customers they are talking to a machine. The obligation stems from Article 50 of the EU's AI regulation and covers virtually every channel of contact with an algorithm, from a chat window on a website to a bank's customer service hotline.

What exactly must be disclosed

The regulation isn't limited to chat windows on websites. The European Commission explicitly lists automated phone consultants, virtual advisors in banking apps, bots guiding customers through complaint processes, and systems that respond on behalf of public authorities. If a language model, not a live employee, is on the other end, the customer must be told before continuing the conversation.

The form of the disclosure should match the channel. In a chat window, a short sentence at the start of the conversation is enough. On a phone call, it should be a voice message played before the system starts answering questions. On voice-controlled devices, the information should be given before the actual interaction begins.

Deepfakes and content of public interest

The rules go further than chatbots alone. Realistic video material, manipulated voices, and algorithm-generated images will require permanent watermarks or visible labels disclosing their origin. A separate category covers content on public affairs, such as articles about politics, health, or the economy, where artificial intelligence played a key role in producing the text. These too must be marked as generated or substantially assisted by AI.

The obligations mainly apply to corporations, marketing agencies, news outlets, and large social media platforms. Ordinary users publishing privately generated content are not subject to the same requirements, which sets this rule apart from a general ban on disinformation.

What it means for companies in Poland

For Polish businesses, this is the next stage in rolling out the AI Act, following earlier obligations for high-risk systems. The practical problem is that many companies still lack a full inventory of where artificial intelligence actually operates within their processes, from banks to public offices using automated hotlines. Without that inventory, it's hard to assess which customer touchpoints even need to change before the August deadline.

The European Commission has prepared a code of practice, a set of guidelines to help companies implement the rule without misinterpreting it. Some commentators compare the scale of the change to the rollout of GDPR or cookie notices, since it affects almost every company with online customer contact, not just AI technology providers.

Penalties and enforcement

The highest fine threshold, €15 million or 3 percent of global turnover, puts the transparency obligation in the same league as the most serious AI Act violations, alongside requirements for high-risk systems. For large platforms and banks, this represents real financial risk if a customer encounters a bot without a clear disclosure of who they're talking to. National supervisory authorities in member states, including Poland, will be responsible for enforcing the rule locally.

The August 2, 2026 deadline leaves companies little time to adjust their processes, especially where automated customer service was rolled out gradually and without a consistent disclosure standard. Lawyers specializing in the AI Act expect a wave of internal audits in August and September, as companies begin actually checking where the required labeling is missing.

Sources: Forsal.pl (forsal.pl), Dziennik.pl (dziennik.pl)