AI labelling: what changes for sales emails and chatbots

In short: a website chatbot must say it is AI, realistic AI images and video must be labelled, but on the wording of the Act a one-to-one sales email needs no specific label. 2 December is the marking deadline for AI tool providers. Here is what it means for a small business.

The EU AI Act's transparency rules have applied since 2 August 2026, and the last transitional deadline falls on 2 December: by then, providers of generative AI systems that were already on the market must mark AI-generated content in a machine-readable way. For a small business this can sound like someone else's problem, since we don't build the large language models ourselves. It is still worth checking now where your company uses AI in front of customers: the website chatbot, sales emails, marketing images and videos. Below we sum up what the regulation actually says, what is likely to apply to a smaller company, and a short checklist. This is not legal advice; for a specific case, ask a lawyer.

1. what changed, and when?

According to the European Commission's AI Act Service Desk, transparency obligations apply and are enforceable from 2 August 2026. The exception is Article 50(2), the machine-readable marking of AI-generated content: providers whose systems were placed on the market before 2 August have until 2 December 2026 to comply. That four-month transitional period was introduced by Regulation (EU) 2026/1744, the so-called Digital Omnibus on AI.

The same amendment pushed most high-risk rules back: 2 December 2027 for Annex III systems and 2 August 2028 for AI embedded in regulated products. Most AI tools used in sales and marketing are not high-risk, so for small and mid-sized companies the transparency rules are the part that matters right now.

2. provider or deployer?

The Act splits obligations between two roles. A provider develops an AI system, or has one developed, and places it on the market or puts it into service under its own name. A deployer uses an AI system under its authority in a professional context. A company that writes emails or generates images with an off-the-shelf tool is usually a deployer. If you build and run a chatbot under your own brand, some provider obligations may land on you, so clarify this with whoever builds it.

3. what it means for sales emails

For text, Article 50 asks deployers to disclose AI generation in one narrow case: text published to inform the public on matters of public interest. Even then there is an exemption when the content has gone through human review or editorial control and someone holds editorial responsibility.

A tailored proposal, a quote follow-up email or a meeting recap does not fall into that category on the wording of the Act, and machine-readable marking is the AI tool provider's job, not the sender's. Two habits are still worth keeping. Keep a person in the loop: we put every outgoing email in front of a human for approval, because responsibility stays with the sender. And don't strip or work around marking that a tool has built in when you pass content on.

4. website chatbots must say they are AI

Article 50(1) requires AI systems that interact directly with people to be designed so that people know they are talking to an AI, unless that is obvious to a reasonably well-informed, observant user. Article 50(5) adds that this must be clear and given at the latest at the first interaction.

In practice, that is one visible line at the top of the chat window, such as "An AI assistant is answering; ask any time to reach a colleague." The same logic applies to AI phone assistants and automated support channels that might pass for a person. Formally it is the provider's duty, but customers see your brand, so show it even when the chatbot comes from a vendor.

5. images, video and voice: the deepfake rule

If AI-generated or edited image, audio or video content resembles real people, objects, places or events and could falsely appear authentic, the Act calls it a deep fake, and the deployer must disclose that it is artificial. Think of an AI voice imitating a real colleague, a "photo" of an event that never happened, or a generated product shot that looks better than the real thing. For evidently artistic, creative or satirical work, disclosure can be made in a way that does not spoil the work.

6. how big is the risk?

Under Article 99, breaching Article 50 can mean fines of up to EUR 15 million or 3% of worldwide annual turnover, whichever is higher. For SMEs and start-ups, the cap is whichever of the two is lower. These are ceilings, not automatic fines, but they show labelling is more than a courtesy.

AI labelling checklist for small businesses

  • List every place your company uses AI in front of customers: chatbot, emails, images, video, voice.
  • For each tool, decide whether you are the provider or only a deployer.
  • Chatbots and voice assistants: a visible "AI is answering" line in the first message.
  • Marketing images and video: label AI content that resembles real people, products or events.
  • Sales emails: a person approves sending; never remove a tool's built-in marking.
  • Ask your AI vendors how they meet the 2 December marking deadline.
  • Name the person responsible for reviewing AI content, in writing.

frequently asked questions

do we have to label a proposal written with AI?

On the wording of the Act, there is no specific deployer labelling rule for a one-to-one business email; text disclosure covers content published to inform the public on matters of public interest. You remain responsible for what you send, which is why human review is standard practice for us.

what happens on 2 December 2026?

The transition ends for generative AI providers whose systems were on the market before 2 August: from then on they must mark generated output in a machine-readable way. New prohibitions on generating non-consensual intimate material and child sexual abuse material also apply from that date.

what about the high-risk rules?

After the amendment they apply from 2 December 2027, and from 2 August 2028 for AI embedded in regulated products. Most sales and procurement AI tools fall outside that scope, but check your own case.

transparent AI for your customers

We build sales, procurement and email AI agents so that a person approves every outgoing step and the chatbot says it is AI in its first line. To review your own workflow, read our guide to AI sales meeting preparation, see what we do on the homepage, or talk to oiSystem.

sources