DarkVision

Acceptable use

Updated 18 September 2026. These rules form part of the Terms of use. They set out what the service may not be used for, how that is enforced, and where responsibility lies. Using the service means you have read them and agree to them.

1. You are responsible for your use

The service is a technical relay between you and third-party machine-learning models. You alone decide what to write, what to upload, what to generate and what to do with the result. You are solely responsible for all of it and for complying with every law, regulation and third-party right that applies to you, including the laws of the country in which you are located. Nothing on this site grants you permission to do anything, and no absence of a technical restriction is to be read as such permission.

We do not create the models, we do not author their output, and we do not read your conversations. They are stored encrypted with a key derived from your access code; that key is not kept on our servers, and the plain text of a request exists on our side only for the moment it takes to relay it to the model provider, as the Privacy notice explains. The output is generated automatically in response to your instructions. It may be inaccurate, offensive or unlawful to use in your jurisdiction; evaluating and acting on it is your decision and your responsibility.

2. Prohibited uses

Four categories below — sexual content involving minors, planning or supporting terrorism, and building or using nuclear or chemical weapons — are also refused by an automatic check before a request is forwarded anywhere. A refused request is not sent, not forwarded and not charged. Everything else on this list is prohibited whether or not anything stops it.

You must not use the service, or anything obtained through it, to:

  • Break the law. Any use that is unlawful where you are, where we operate, or where the effect of your conduct is felt is prohibited, whether or not it is listed below.
  • Sexualise minors. Any sexual or sexualised depiction of a person under 18, real or fictional, in any medium or framing, including roleplay, and any assistance with grooming or with the production or distribution of such material. Refused automatically.
  • Plan or support violence. Planning, facilitating, financing or recruiting for terrorism, attacks on people or property, or any other act of violence; seeking synthesis routes, construction or dispersal methods for chemical, biological, radiological, nuclear or explosive weapons. Terrorism and nuclear and chemical weapons are refused automatically; the rest is prohibited all the same.
  • Harm a real, identifiable person. Harassment, stalking, threats, doxxing, defamation, intimate or sexual imagery of a person without their consent, impersonation, or any content designed to deceive others about a real person.
  • Defraud or deceive. Scams, phishing, forged documents, fraudulent impersonation of organisations, and material intended to mislead the public.
  • Compromise systems. Creating or deploying malware, gaining or attempting unauthorised access to systems or data, or attacking any network or service, including this one.
  • Infringe others' rights. Reproducing or distributing material in violation of copyright, trademark, privacy, publicity or other rights.
  • Abuse the service. Circumventing rate limits or the automatic checks, sharing or reselling access, payment fraud, and any interference with the operation of the service or other users' use of it. For partners who hand out promo codes: earning commission through accounts you control, through several accounts, browsers or addresses of one person, or through misleading or unsolicited promotion.

3. What the absence of screening does and does not mean

In the AI section, apart from the automatic checks described above, we do not pre-screen, read, rate or classify your requests. Stored conversations are ciphertext to us without your code, and while a request is being relayed to the model provider nothing of it is logged or looked at. That is a matter of how the service is built and run, not of end-to-end encryption: the provider of the model necessarily sees the text it is asked to answer. None of that is a licence. The prohibitions in section 2 apply in full whether or not anything stops a particular request, and a breach is a breach even if the request went through.

The service does not prohibit lawful content solely because it is adult, uncomfortable or controversial. Whether particular content is lawful for you is determined by the law that applies to you, and only you can make that assessment. The models are third-party software and may decline a request on their own; that is their behaviour, not a rule of ours.

A language model is not a person. If you are in crisis, or the subject matter concerns your own safety, please contact a local emergency or crisis line.

4. Enforcement

Because an access code is all we know about you, suspending or closing the code is our principal sanction. We act on the automatic check, on substantiated reports from affected people, on lawful demands from competent authorities, and on technical abuse signals. A code closed for breach of these rules loses any remaining balance where the law permits. We may also preserve and disclose the limited data we hold — payment records and timestamps, never the content of stored conversations, which we cannot decrypt without your code — where the law requires it.

Locks placed by the automatic guard are temporary and lift on their own, each repeat lasting longer than the last; we may lift one earlier or, where the pattern leaves no doubt, make it permanent. A locked code can always sign in, read, clear and delete its own data.

Misuse reports, legal or rights complaints, and anything else that needs a record: contact@darkvision-sec.com. Questions about these rules may also be asked in our Telegram channel, @darkvisionsec, which accepts direct messages; a report made there is not a formal notice, and we will ask you to send it by email. Include enough detail for us to identify what you are describing; we cannot search conversations to find it for you.