Under Article 50(1) of the EU AI Act the duty to disclose follows whoever builds the AI system, not the brand that speaks through it. Razviel builds it, so Razviel carries it. This page states what the system does, and which obligation each part answers to.
Under the EU AI Act, the duty to disclose follows whoever builds the AI system, not the brand that speaks through it. Razviel builds it, so Razviel carries it.
Swipe
Each marker is one law, placed on the territory it governs. What each requires, and what answers it, is below.
Every medium keeps identity in a different place — a header, a sign-off, the message itself. The disclosure is designed to sit where that medium already answers who wrote this.
The identification is applied by the system rather than left to the model to remember: the design does not send a first message that lacks it.
Every medium keeps identity in a different place. The disclosure is designed to sit where that medium already answers who wrote this.
The label follows each message, not the account. A message written by the AI identifies the AI; a message written by a person carries that person’s name.
The label follows each message, not the account. If a customer asks whether they are speaking with AI, the answer is yes.
Five named provisions across three legal territories. Four of them impose the same shape of duty — tell the person, at or before the first interaction. One imposes a different one: answer truthfully when asked.
| Rule | What it requires | Status |
|---|---|---|
| EU AI Act, Article 50(1) European Union | Providers must design and develop AI systems that interact with people so that those people are informed they are interacting with an AI system, in a clear and distinguishable manner, at the latest at the time of the first interaction. | by design |
| Maine, 10 M.R.S. §1500-DD United States · Maine | A chatbot may not be used in trade or commerce in a manner that may mislead a reasonable consumer into believing they are engaging with a human being, unless the consumer is notified clearly and conspicuously. | by design |
| California, B&P Code §17941 United States · California | Where a bot is disclosed, the disclosure must be clear, conspicuous, and reasonably designed to inform the person that they are communicating with a bot. | by design |
| DMCC Act 2024 United Kingdom | Material information a consumer needs must not be omitted, and a consumer must not be misled about who or what they are dealing with. | by design |
| Utah AI Policy Act United States · Utah | On request, a person must be told whether they are interacting with generative AI rather than a human. | in force |
This page names the specific obligations Razviel addresses and states what the system does about each one. It does not assert general conformity with any regulation as a whole, and Razviel holds no certification for AI transparency — none exists for Article 50, which has no conformity assessment or marking outside high-risk systems.
Four impose one shape of duty: tell the person, at or before the first interaction. One imposes a different one: answer truthfully when asked.
Article 50(1) attaches to whoever designs and develops the AI system, and that is Razviel. An assistant that cannot say what it is creates a problem where it was built, not where it was deployed. Your own consumer-facing compliance remains yours — that has not changed, and your agreement says so — but this particular duty is met in the product before it reaches you.
Nothing here asks you to take a vendor’s word. Both clauses are published in full on this site and bind Razviel on the day you sign, the obligations are named by article rather than described in general terms, and the identification appears in the conversation itself. Your counsel can verify all of it without a call.
Replies go out in your brand’s name and voice on every channel. The identification says the sender is AI and does not interrupt the answer to do it. The one place Razviel appears is a quiet “Powered by Haven · Razviel” at the foot of the live chat.
None of this is a courtesy. It is the condition on which a support AI can be operated in someone else’s name.
The same rule appears in two published documents on this site — one written to the party that signs, one written to the person on the other end of the conversation.
Razviel’s automated agents are designed never to assert that they are human, and to answer truthfully when asked whether they are AI.
Our automated agents never claim to be human. They will not tell you they are a person, and they will not invent a human colleague or supervisor. … If you ask whether you are talking to AI, you get a truthful answer. Always.
Two published documents carry the same rule: the automated agents are designed never to assert that they are human, and to answer truthfully when asked.
Written the way a reviewer asks them, in the order they usually arrive.
Under Article 50(1) the duty sits with the provider — whoever designs and develops the AI system. That is Razviel, and it is not something a contract can move to you: a brand deploying a support AI carries no chatbot-disclosure duty of its own under that Article. You remain responsible for your own consumer-facing compliance generally, as your agreement sets out, but the identification itself is built into the product rather than left for you to switch on.
Email and Live Chat today, every channel as it connects. Where the surface belongs to Razviel there is a label in the conversation header; where it does not, the identification is carried in the message itself — the sign-off in email, and the first message of a conversation on Instagram, Facebook, WhatsApp, TikTok and SMS.
No. It is part of how the system is built rather than a setting, for the reason above: the obligation is Razviel’s, so it cannot depend on a configuration choice made in your account.
No, and this is contract text rather than a design intention. Razviel’s automated agents are designed never to assert that they are human, never to invent a human colleague or supervisor, and to answer truthfully when asked whether they are AI. Both the Services Agreement and the Privacy Policy say so, and both are published on this site.
The label follows each message, not the account. A message written by a human carries that person’s name; a message written by the AI identifies the AI. A permanent account-level badge would be inaccurate half the time on a shared inbox, which is why the design does not use one.
Yes. The identification is designed to be rendered in the language the conversation is being held in, since an English label shown to a German customer would not inform them.
In the live chat only, and quietly. Replies speak in your brand’s name and voice on every channel, and the identification says the sender is AI. The foot of the live chat reads “Powered by Haven · Razviel”; email, social and SMS replies carry your brand alone. Where your own privacy disclosures need to name Razviel as a processor, that is handled in your Data Processing Addendum.
No, and no such certification exists. Conformity assessment and CE marking under the EU AI Act reach high-risk systems; Article 50 transparency has no assessment, no notified body and no mark. What this page offers instead is the specific obligation named, what the system does about it, and the contract text that is already in force — all of which you can check.
Both clauses are published in full: Services Agreement §14 and Privacy Policy §5. Data handling and the split of responsibilities are in the Data Processing Addendum.
A question this page does not answer? Write to legal@razviel.com and the answer is added here.
What a brand usually asks first.
Under Article 50(1) the duty sits with the provider — whoever designs and develops the AI system. That is Razviel, and it is not something a contract can move to you: a brand deploying a support AI carries no chatbot-disclosure duty of its own under that Article. You remain responsible for your own consumer-facing compliance generally, as your agreement sets out, but the identification itself is built into the product rather than left for you to switch on.
Email and Live Chat today, every channel as it connects. Where the surface belongs to Razviel there is a label in the conversation header; where it does not, the identification is carried in the message itself — the sign-off in email, and the first message of a conversation on Instagram, Facebook, WhatsApp, TikTok and SMS.
No. It is part of how the system is built rather than a setting, for the reason above: the obligation is Razviel’s, so it cannot depend on a configuration choice made in your account.
Yes. The identification is designed to be rendered in the language the conversation is being held in, since an English label shown to a German customer would not inform them.
In the live chat only, and quietly. Replies speak in your brand’s name and voice on every channel, and the identification says the sender is AI. The foot of the live chat reads “Powered by Haven · Razviel”; email, social and SMS replies carry your brand alone. Where your own privacy disclosures need to name Razviel as a processor, that is handled in your Data Processing Addendum.
No, and no such certification exists. Conformity assessment and CE marking under the EU AI Act reach high-risk systems; Article 50 transparency has no assessment, no notified body and no mark. What this page offers instead is the specific obligation named, what the system does about it, and the contract text that is already in force — all of which you can check.
A question this page does not answer? Write to legal@razviel.com and the answer is added here.
Swipe