Ethical AI at Xminds

• Interpretation, not impersonation. Our mentors are clearly labelled AI interpretations of historical figures based on public sources — never a claim of authenticity, endorsement or affiliation.
• Public domain only. We do not build personas of living people, and no likeness or voice is cloned from a private individual.
• Education over engagement. The product is designed for reflection and learning, not for parasocial attachment, romantic role-play or emotional dependency.
• Transparency by default. Every surface of the product states that you are talking to an AI.
• Human accountability. A named operator (see the imprint) is responsible for what is published; every persona passes a human review before it goes live.

Before a mentor can be created, the persona passes an automated rights review that blocks the request outright when any of the following apply:

• the person is, or may be, still living — personality rights apply and no public-domain term has begun;
• the person died less than 70 years ago (the widely used life + 70 term for copyright and neighbouring rights);
• the estate is known to license name, image and likeness commercially — such estates are on a hard blocklist regardless of the death year.

Personas that entered the public domain only recently are flagged as elevated risk and are additionally restricted from reproducing long passages of protected works. Blocked requests are logged with the reason and are visible to our review team; they cannot be overridden in the normal publishing flow.

Portraits, intro videos and voices are generated assets, produced for Xminds in a period-appropriate artistic style. They are not photographs, recordings or biometric clones of any person, and they are not scraped from rights-managed archives.

Mentor personas are grounded in publicly available biographical and historical material, and each persona is reviewed by a human before publication. Quotations shown on mentor pages are attributed and drawn from public-domain works; conversational output is generated and is not presented as verbatim quotation.

AI output can still be wrong, outdated or invented. We say so on the product surfaces where it matters, and we ask you to verify anything consequential against primary sources. Nothing in Xminds is medical, legal, financial or psychological advice — see our Terms of Service.

• System prompts constrain every mentor to an educational, respectful register and prohibit hate speech, harassment, sexual content, self-harm encouragement and instructions for illegal or dangerous acts.
• Provider-side safety filters run on top of our own instructions for text, speech and video generation.
• Conversations that suggest acute distress are steered toward real-world help rather than continued role-play.
• We may filter or refuse outputs, remove content and suspend accounts that abuse the service; the rules users agree to are in the acceptable-use section of our Terms of Service.
• Errors are captured, grouped and reviewed continuously so that recurring quality or safety issues are fixed rather than repeated.

• No personas of living people, public figures included.
• No voice or face cloning of private individuals, and no deepfakes presented as real footage.
• No claim that a mentor speaks for the real person, their descendants or their estate.
• No political campaigning, no medical or legal advice, no adult content.
• No sale of personal data, and no training of third-party foundation models on your conversations.

You can use Xminds as a guest. With an account, your conversations are stored so you can return to them, and you can delete them or your whole account at any time in settings. Consent for analytics and marketing cookies is opt-in in the EU/EEA, the UK and Switzerland, and can be changed at any time from the footer. Full detail, including our processors, is in our Privacy Statement.

Xminds is operated from Germany under EU law. In terms of the EU AI Act, the service is a limited-risk generative AI application whose central obligation is transparency: users are told they are interacting with AI and that content is AI-generated. We do not operate any prohibited practice, we perform no biometric identification, emotion inference or social scoring, and the service is not used for any high-risk purpose in the meaning of Annex III.

Payments are processed by our reseller Paddle.com as Merchant of Record, which also handles invoicing, tax and refunds. We do not store card data.

If you believe a persona, quotation, image or video infringes a right, is factually wrong, or crosses an ethical line, write to hello@xminds.app with the page URL and a short description. We review every notice, act on justified claims — including removing or unpublishing a mentor — and we will tell you the outcome. Descendants and estates can request removal without having to prove a legal claim first.