A customer calls to check their account balance. Another sends a WhatsApp message to change a hotel booking. Someone else wants to schedule an appointment at a car repair shop. In all three cases, it is increasingly likely that the response will come not from a person, but from a virtual assistant.

And since 2 August 2026, there is a legal obligation that many companies have yet to fully address: customers need to know when they are interacting with AI.

This requirement comes from Article 50 of the European Union’s Artificial Intelligence Act (AI Act). In our conversations with clients across different industries, we repeatedly come across the same misconceptions. Let’s debunk them one by one.

Myth 1: “I use AI to assist people, not replace them, so this doesn’t apply to me”

This is often the first assumption, and it is a mistake.

The regulation does not distinguish between “AI that replaces people” and “AI that helps provide better customer service”. What matters is one thing: if an AI system interacts directly with a customer, the customer must be informed that they are interacting with a machine, unless this is obvious from the context.

It does not matter if the chatbot only identifies the reason for the enquiry before transferring the customer to a human agent. If it interacts with the customer, it counts.

Myth 2: “The legal department can solve this by adding a clause”

Not quite.

Article 50 is not something that can be addressed simply by signing a document once and considering the matter closed. It is a requirement that needs to be embedded in the customer experience itself: in the welcome message of a voicebot, in the first message of a chat conversation, and in the way a WhatsApp assistant introduces itself.

In practice, this means reviewing scripts, conversational flows and automatically generated content, such as call summaries or semi-automated email responses, that reaches customers without making it clear that AI has been involved in producing it.

Myth 3: “If my BPO provider manages the channel, the responsibility is theirs”

It depends, and this is where many operations are caught by surprise.

The responsibility for informing customers lies with the party deploying the technology in front of the end customer. If you outsource customer service operations, this responsibility should already be clearly reflected in your contracts and SLAs with your BPO provider, rather than simply being assumed.

What Does This Look Like Across Different Industries?

Managing this requirement in banking is not the same as managing it in tourism. Here are some examples of where problems commonly arise:

• Banking: virtual assistants handling balance or transaction enquiries, where the initial AI disclosure may be buried within an options menu.
• Travel and tourism: chatbots managing bookings and travel changes, particularly during peak periods when the bot may handle the entire conversation without clearly identifying itself as AI.
• Automotive: workshop appointment assistants that often rely on legacy scripts created before this obligation existed.
• Healthcare and pharmaceuticals: bots used to schedule appointments or answer basic queries, a particularly sensitive area given the nature of the information involved.

The same pattern appears across all these industries: the technology was originally deployed to reduce waiting times, while the “I’m a virtual assistant” disclosure was treated as a minor design detail.

The Real Issue Is Not the Fine

Penalties exist, but what we see in many operational audits is a more fundamental problem: many companies do not have a clear inventory of every customer touchpoint where AI is interacting with customers.

Without that map, there is no way to guarantee compliance with the regulation, no matter how good the company’s intentions may be.

Before December Arrives

Generative AI systems that were already in operation before August 2026 have until December to adjust their technical labelling. However, this deadline should not be confused with the obligation to inform users, which is already enforceable.

There are two questions we usually ask when we begin working with an operation:

Have you mapped every touchpoint where AI interacts with your customers?

Do your contracts with BPO providers clearly define who is responsible for meeting this obligation?

If you do not have a clear answer, this is often the starting point for our CX Consulting and AI Technology projects at MST Holding: identifying where AI is being used across your operation, how this is communicated to customers, and how it is documented.

Do You Know Whether Your Operation Falls Within the Scope of Article 50?

Talk to our CX Consulting and AI Technology team and find out how to assess your customer service operation against the requirements of the EU AI Act.