Artemii
Sign in Open Artemii
Legal

Artemii and AI: what you should know

Draft pending legal review. This is our current wording, written from how Artemii actually works. A solicitor has not yet reviewed it, and it may change. Mayfair's company details will be added once confirmed.

Last updated 5 November 2026. Artemii is made by Mayfair.

This page explains, in plain words, that Artemii is an AI, what it can get wrong, how AI-made content appears, and what control you have.

You are talking to an AI

Artemii is an artificial intelligence system made by Mayfair. Every reply in chat, in the Chrome extension, in the embedded assistant and through the developer API is written by an AI model, not by a person. Mayfair staff do not write or review replies in real time.

  • The models are instructed never to claim to be human and never to deny being an AI. If you sincerely ask, they say plainly that they are an AI. This holds in Direct mode and for custom assistants and API callers: instructions written by an assistant author or an app cannot change identity, honesty or safety rules.
  • Arlo, Lucian, Orion and Curzon are model names, not people.
  • The Terms of Use say that Artemii generates text using statistical models and can be wrong.

What the AI can and cannot do

  • It can be wrong, out of date, incomplete or biased, and it can state falsehoods confidently, including invented facts, quotes, sources and links.
  • It has no live knowledge unless web tools are on. When they are off or unavailable it is told to say so.
  • It is not a lawyer, doctor or financial adviser. Do not rely on it for legal, medical, financial or safety decisions. In Direct mode it gives fewer warnings, but the same limits apply.
  • It can be tricked by malicious text in web pages, files or tool results. Artemii labels such text as untrusted and asks you before actions that change something, but this is a safeguard, not a guarantee.
  • It does not see pictures you upload. It only knows that a file was attached.
  • Replies may be similar for different users and may resemble other people's work.
  • Details for each model are on the Model cards page. None of the models has been independently benchmarked.

Content made by AI

  • Documents, web pages, designs, code and spreadsheets the assistant produces are shown as artifacts in a panel beside the chat, clearly separate from your own input.
  • Pictures are made by an AI picture model and stored as produced. Artemii does not currently add a visible label, watermark or embedded provenance mark to generated pictures, to text, or to shared pages.
  • You are responsible for how you use and present AI output. The Acceptable Use Policy forbids passing AI output off as human where that would mislead, and forbids sexual images of real people (those picture requests are refused).
  • Direct mode is for adults aged 18 or over. People who use it agree not to present AI content as someone else's words and not to share Direct mode content with anyone under 18.

Your information and control

  • Memory: see, edit, export and delete everything Artemii remembers, switch memory off, or switch automatic learning off.
  • Sources: when web tools are used, the answer shows the sources it cites and what it searched or read.
  • Tool use: a log of the tools, apps and connections Artemii used is in Settings for 90 days.
  • Approvals: actions that change something in a service or app ask first by default.
  • Human oversight: you can stop any answer, decline any action, switch off tools, disconnect services, revoke everything in one click, and use incognito chats.
  • Your data: see the Privacy Policy and Data retention and deletion.

How Artemii is built

  • Artemii runs open-weight language and picture models on GPU servers that Mayfair rents and controls, with Mayfair's own instructions, tools and safety settings. Credits and licences are on the Licences and model notices page.
  • Arlo, Lucian, Orion and Curzon are configurations tuned for different tasks.
  • Mayfair did not build the underlying models from scratch and does not control their training data.

Decisions about people

Artemii is a general assistant. It does not make decisions about you with legal or similarly significant effect, and the Acceptable Use Policy forbids using it for such decisions about other people without human review. The one automatic action against an account is a temporary pause (about 15 minutes, doubling on repeat up to 24 hours) when the service sees rapid repeated rate-limit refusals. Reading, exporting and deleting your data keep working during a pause, and an administrator can lift it.

Report a wrong or harmful answer

Email hello@mayfairmail.com with "AI answer" in the subject, with what you asked and what Artemii said.