European AI regulation is moving from theory to practice: Article 50 of the EU AI Act became legally binding on August 2, and the European Commission has already begun enforcement. In the week of August 12–13, the Commission's AI Office and national regulators started actively monitoring the new transparency rules. In concrete terms: every chatbot, AI agent, and virtual assistant operating in the EU must clearly inform users they're communicating with a machine – unless that's already obvious.
The essentials
- In force now: Article 50 of the EU AI Act is legally binding as of August 2, 2024; regulators began enforcement on August 12–13
- Penalties: Fines up to €15 million or 3% of global annual turnover (whichever is higher); startups face the lower figure
- Who must comply: All providers of chatbots, AI agents, virtual assistants, and AI-generated content (images, audio, video, deepfakes) serving EU users
- Enforcement capacity: The Commission has assembled a 38-person team in Brussels with power to interview staff, demand documentation, and ultimately remove products from the EU market
What exactly is required now?
The rule is straightforward but strict: a chatbot must clearly state at the point of interaction – where the user actually engages with it – that they're communicating with AI. Hiding this disclosure in terms and conditions doesn't count. A watermark alone is insufficient.
For AI-generated content (images, audio, video), a visible or machine-readable label is required. Deepfakes must be flagged – regardless of whether deception was intended.
One exception applies: the machine-readable marking requirement for AI-generated content has a grace period until December 2026 for tools already on the market before August 2. Everything else takes effect immediately.
Who signed – and who didn't
The industry was partially prepared: more than 180 companies have already signed the EU Code of Practice on AI transparency – including Anthropic, OpenAI, Google, Microsoft, Amazon, IBM, Mistral AI, and Cohere. This signature carries practical weight: regulators treat signatories as acting in good faith and are more lenient in case of disputes.
Meta did not sign the code, citing legal uncertainty. This is a disadvantage: without this "goodwill buffer," Meta will likely face stricter regulatory scrutiny.
Tight timelines for industry
The Commission gave companies minimal preparation time: final guidelines were published just two weeks before the deadline. Chatbot providers had to redesign disclosure flows, add banners, and rewrite documentation – in a sprint, not a marathon.
For large platforms, the penalties are substantial. For startups, they can be existential, though the Commission has committed to assessing smaller companies against the lower of the two penalty options.
What this means for German companies
Any German company operating a chatbot or distributing AI-generated content in the EU must act now – or face significant fines. Critical point: disclosure must occur at the interaction point itself, not hidden in terms of service. Companies that haven't yet should consider whether the EU Code of Practice is relevant to their business model; signing could reduce regulatory pressure. And: the 38-person enforcement unit in Brussels is not symbolic – it has real authority.
Sources
Editorially owned by Ideal Syka. Sources and method: Newsroom & method. Tips and corrections: ai@i6eal.de.
