AI transparency

Last updated: 01 August 2026

You are talking to an AI

The assistant on this website is an AI system, not a person. We tell you this because Article 50 of the EU AI Act requires it, and because you should know regardless.

• It answers using a language model running on hardware we own and operate in Sofia.
• It is not a person, and it is not a person following a script.
• Nobody reads your messages in real time unless the conversation is escalated to us.

What it can do

• Answer questions about what we do, how we work, and what things typically cost.
• Book a 30 minute diagnostic call in our calendar.
• Hand the conversation to a person and take your email so we can reply.

What it cannot do

• It cannot commit us to anything. It cannot agree a price, sign anything, or form a contract. Nothing it says is a binding offer.
• It cannot see your account, your files, or anything you have not said in the conversation.
• It can be wrong. It is a language model, and language models sometimes state incorrect things confidently. Anything that matters commercially should be confirmed with a person before you rely on it.

Please do not send it

Personal data about other people, passwords or access credentials, client confidential material, or anything you would not put in a first email to a supplier you have not met.

What we keep

By default the conversation is answered in memory and not stored. It is kept only if you consent, and then for 24 months from your last activity. To block abuse we keep a salted hash of your IP address for 24 hours, never the address itself. Full detail is in the privacy policy.

The assistant classifies conversations internally to route them and to decide when to offer a call. That routing does not produce a decision with legal or similarly significant effect about you.

Reaching a person

Ask the assistant for a human and it will take your email and pass it on. You can also write to office@cloudsource.bg or call +359 888 044 260.

Systems we build for clients

The same principle applies to the systems we build. People interacting with them are told they are interacting with AI, permissions are enforced at the tool level rather than in a prompt, and a route to a human always exists. Where we are the provider of an AI system and the client is the deployer, we set out which obligations sit with whom in the engagement contract rather than leaving it to be discovered later.