EU AI Act: from August 2 your chatbot has to identify itself — what this means for small businesses

August 2, 2026 passed largely unnoticed, but on that date part of the EU AI Act formally became binding: the transparency obligations under Article 50. Unlike much EU regulation, there’s no size threshold — the obligation applies to every business that offers or deploys AI systems toward users in the EU, from multinationals to a one-person business with a chatbot on their website.

What actually changed

The core of Article 50 is simple to state: if a customer is communicating with a chatbot or voicebot, that has to be clear to them. Not a buried disclaimer, but recognizable, timely information that they’re not talking to a human. There’s also a labeling obligation for AI-generated text, images and audio you publish, and an always-applicable disclosure requirement for deepfakes. A related obligation — that everyone working with AI within a company must have sufficient understanding of its capabilities and risks (“AI literacy”) — has formally applied since February 2025, and now has to be demonstrable, not just present.

At the same time, some nuance is needed: the heavier obligations for so-called high-risk AI systems were actually delayed. Through the Digital Omnibus on AI, which entered into force on July 27, 2026, the Annex III obligations shift to December 2027 and the Annex I obligations to August 2028 — with the explicit goal of reducing the compliance burden, by as much as 35% for small and medium businesses by 2029. The Article 50 transparency obligation was not part of that delay, and so it applies as planned since August 2.

What this means if you’re still running on Excel and Outlook

For a small business owner who recently put a chatbot on their website, uses an AI writing assistant for customer communication, or is considering working with AI-generated content, this isn’t a distant concern. A chatbot that doesn’t identify itself as such is immediately non-compliant — regardless of whether it’s a simple widget from an external vendor or something built in-house. The same applies to AI-generated images in marketing materials or automatically drafted text presented as human work.

The fine side, and the trust side

Violations of the transparency obligations can result in fines of up to €15 million or 3% of global annual turnover, whichever is higher — in practice a theoretical ceiling for most small businesses, but still a signal that regulators take this seriously. More important than the fine may be the other side of the coin: a customer who knows upfront that they’re talking to an AI system trusts that interaction more than one who only finds out afterward. Transparency isn’t just a box to tick — it’s also a way to preserve trust at a moment when customers are increasingly running into AI without immediately realizing it.

One concrete first step

This week, put together a simple overview — an AI register — of every AI tool your business uses toward customers: a chatbot, an AI writing assistant for emails, a tool that generates summaries or content. Check for each one whether it’s clear to the customer that it’s AI. That overview takes an afternoon, prevents you from overlooking an obligation, and is immediately the starting point for the internal AI policy the law also asks for.

Sources

• Score Agency, “EU AI Act: deze verplichtingen gelden vanaf 2 augustus 2026”
• ActCheck, “EU AI Act voor het MKB: wat moet u doen vóór 2 augustus 2026?”
• ISMS.online, “The EU AI Act August 2026 Deadline Has Been Delayed: What it Means for Businesses”
• Travers Smith, “EU agrees to delay key AI Act compliance deadlines”