PIPELINEGuide
EU AI Act: what small businesses must actually do in 2026
The regulation sounds like a corporate issue, but it affects every business with a chatbot on their website. What's actually required — without panic and without downplaying.
Since August 2026, the transparency obligations under Article 50 of the EU AI Act apply to every business with a chatbot on their website, regardless of company size: anyone communicating with an AI system must be able to recognize it. For most small businesses, a manageable package is sufficient — an inventory, labeling, a brief usage policy, and documented training.
01
What applies since August 2026
The transparency obligations under Article 50 are in effect. Core principle: anyone communicating directly with an AI system must be able to recognize it. A chatbot must disclose upon first contact that it is not a human. Published content that has been synthetically generated or substantially altered must be labeled.
02
Who is affected
The AI Act does not distinguish by company size. A sole proprietorship with a website chatbot is just as affected as a corporation. What matters is the purpose of deployment and the associated risk, not the number of employees.
03
AI competence
Companies that deploy AI systems must ensure that the people working with them have sufficient competence. This is not an examination requirement, but it should be documented — a training session with an attendance list is the simplest proof.
04
What is usually not affected
Purely internal use with no external impact: research, drafts, summaries. Once results are published or customers communicate directly with the system, the situation needs to be reassessed.
05
The practical approach
Take inventory of which tools are used in the business. Classify the use cases. Label where necessary. A brief usage policy. Documented training. For most businesses, this is a manageable package, not a project.
06
What's coming in the next few months
The requirements take effect in several stages, and individual detailed questions — such as the exact boundaries for AI-generated content — will only be clarified through guidelines and initial decisions. If you have a written usage policy now, you can adjust it rather than starting from scratch with every clarification.
Questions
Frequently asked questions
Are there penalties?
The AI Act provides for sanctions that can be substantial depending on the violation. For transparency obligations, the penalties are lower than for prohibited practices. For small businesses, the realistic risk is less the fine itself than the competitive disadvantage when clients demand proof of compliance.
Do we have to label every AI-generated text?
For published content that was synthetically generated or substantially altered, yes. A text that was drafted with AI assistance and then editorially revised and owned by a human is assessed differently than an unedited machine-generated text published as-is. The exact boundaries are not yet definitively settled — a transparent approach is the safe route.
Do we need an AI officer?
The AI Act doesn't require that for most businesses. It still makes sense for one person to be responsible — if only to keep the inventory from becoming outdated.
Next
Relevant Services
Next step
Let's talk briefly about whether this makes sense for you.
A twenty-minute conversation is usually enough to clarify whether a project makes sense. If not, we'll say so — that's cheaper for both sides than an offer nobody needs.