AI transparency at Digital Dynamics

Where we use artificial intelligence

In three places:

• The AI assistant on this site. The replies in the chat are generated by artificial intelligence, not written by a colleague. A permanent bar at the top of the chat window says so — it stays there after your first message.
• The free tools. Output produced by AI in the GDPR checker, the website audit, the demo generator, the AI SEO comparison and MailBeautify carries an "AI-generated" label, plus machine-readable marking in the HTML.
• Client projects. When we build a chatbot or a phone AI for someone, the Act treats us as the provider and the client as the deployer. We build the disclosure in before anything goes live.

How we mark machine-generated content

On two levels, because the Act asks for two: one for you, one for machines.

What you see

• A permanent bar at the top of the chat: "You are chatting with an AI assistant, not a human colleague."
• The assistant introduces itself in its first reply as well.
• AI-written sections of the tools carry their own label.

What machines see

• Every AI reply carries data-ai-generated="true" in the HTML.
• API responses send an X-AI-Generated: true header and an ai_generated field.
• The very first event of the chat stream is the disclosure itself, in three languages.

If you would rather talk to a person

• The AI assistant answers questions and calculates prices. It makes no decision about you: it does not assess, score or filter anyone.
• When it prepares a contact request, you approve the details yourself, and a person gets back to you afterwards.
• You can switch to a human at any point: info@digitaldynamics.hu, +36 30 515 7795 or the contact form. The phone is answered by a person.

Phone AI

We build phone AI for client projects; our own number is answered by a person.

• A voice assistant we build says in the opening sentence of the call that you are talking to a machine.
• The speech is AI-generated, so the audio carries machine-readable marking too.
• Clients can adapt the wording, but they cannot switch the disclosure off.

What happens to what you type into the chat

• Replies are produced by a language model running in an EU data centre. Your messages never leave the European Union, and the models are not trained further on your conversations.
• The conversation lives in server memory and is deleted after 30 minutes of inactivity. It is never written to a database.
• The only thing we keep is what you send as a contact request yourself: name, email address, phone number, message.
• Please do not type sensitive data — passwords, card numbers, health information — into the chat. We do not need it to answer you.

For clients: written statement of compliance

If you order a chatbot or a phone AI from us, the Act makes you the deployer and us the provider. The provider obligations are ours, and we put that in writing.

• A written statement of compliance for your contract, citing Article 50.
• Disclosure built into the system and impossible to switch off.
• Machine-readable marking of generated output, with documented settings.

Legal background and complaints