Arkheon terms of use

Arkheon answers with artificial-intelligence models on one shared account. The terms below are ours: they are the same for the site, the bot, the apps and the API. This is not a formality. A request that breaks them puts the service at risk for everyone users, which is why the check happens before a request goes to work.

What is forbidden

You may not send the models requests or material that contain or aim at:

  • violence and crimes against people S1 · violation
  • fraud, hacking, malicious code S2 · violation
  • sexual crimes S3 · violation
  • sexual content involving minors S4 · immediate block
  • defamation S5 · violation
  • dangerous medical, legal or financial advice S6 · refusal, no violation
  • privacy violations, personal data S7 · violation
  • intellectual property infringement S8 · refusal, no violation
  • weapons of mass destruction S9 · violation
  • hate and discrimination S10 · violation
  • self-harm and suicide S11 · violation
  • explicit sexual content S12 · refusal, no violation
  • election disinformation S13 · refusal, no violation
  • abuse of code execution S14 · refusal, no violation

Separately and without exception: any sexual content involving minors, instructions for weapons of mass destruction and explosive devices, planning violence, writing malicious code, fraud schemes and attempts to obtain other people's personal data. Such requests are refused before they reach a model, and the account is blocked at once.

It is also forbidden to get around moderation (rephrasing, encoding, "role-play" wrappers), to hand the account or API keys to third parties for violations, and to use the service for mass automated mailing and spam.

How moderation works

  1. Every request is checked before it reaches a model. The check is automatic: local rules first, then a safety classifier. The text of a request never reaches the model if the check refused it. A refused request is not charged.
  2. The first violation is a warning. You will see which category fired and a link to this page.
  3. A second violation within 30 days — a 24-hour pause. The models are unavailable in every channel: on the site, in Telegram, in the apps and via API. The balance and conversations are kept.
  4. A third violation within 30 days — the account is blocked. Serious violations (categories marked "immediate block") block from the first time.
  5. A refusal on the model's side counts too. If a request passed our check but was refused during processing, it is recorded as a violation — those are exactly the cases that endanger everyone's access.

"Refusal, no violation" categories (for example, explicit sexual content between adults or requests for medical dosages) are refused but do not count as violations: they bring a block no closer.

What we keep about a violation

For a refused request we keep: the date, the channel (site, Telegram, app, API), the category that fired and the first 400 characters of the request text — so an administrator can review an appeal. Images are not kept. Violation records count for 30 days and then stop affecting the account's status.

Appeal

If you think the check was wrong or the block unfair, file an appeal: arkheonchat.com/appeal. Describe exactly what you asked for and why — that speeds up the review. We reply by email or in Telegram within three working days. If the block is lifted, the violations behind it are removed; if it is upheld, the account stays blocked and unused credits are considered case by case.

The answers come from an artificial intelligence

Every answer in Arkheon is written by a model, not a person. Models make mistakes: they invent facts, links, quotes and figures, confuse dates, go out of date and answer the same question differently. That is a property of the technology, not a fault of the service, and it does not go away on expensive models — it is only rarer there.

The answers are not professional advice — medical, legal, tax, financial or investment. Do not make decisions about health, money, contracts or safety on them without checking at the source or with a professional. Do not use the service where a mistake threatens life, health or property.

Under every answer you can see which model wrote it. What to do with the answer is your decision: you are responsible both for the content of your requests and for how you use the result.

Liability and refunds

The service runs on computing capacity it does not fully control: interruptions, replacements of models in the catalogue and refusals during processing are possible. Accuracy, completeness and fitness of an answer for your task are not guaranteed.

Credits are a prepayment. A failed request is not charged: the hold is returned in full. Spent credits are not refunded — the service for them has been rendered, and an inaccurate or unlucky answer is not a defect. If a charge was our fault — a double charge, a service outage, a paid feature that was unavailable — we return the credits or the money. An unspent balance is refunded on request to support, less the costs actually incurred. The full terms, including the limitation of liability, are in the public offer.

Accepting the terms

Current version of the terms: 2026-09-09. By signing up on the site, on the first message to the bot and in the apps you confirm that you accept these terms, the public offer and the privacy policy; the date, version and channel of acceptance are kept in the account. If the terms change, we will ask you to accept them again.

Support

Questions about payment, credits, models and the apps are answered on the page support, by email support@cpl3.dev or in the bot @arkheon_ai_bot.

These terms complement the public offer and the privacy policy.