Acceptable Use Policy
Last updated 16 November 2026 · Version 1.2
Contents
About this policy
This Acceptable Use Policy sets out what you may and may not do with Kalsenti's website, dashboard, chat feature and API (the "Service"). It forms part of our Terms of Service and applies to everyone who uses the Service, including the end users of business customers' products.
If you break this policy, we may block requests, suspend or close your account, and report illegal activity to the authorities (section 8). Words defined in our Terms of Service have the same meaning here.
1. Always prohibited
These rules apply on every model, including abliterated models, and on every plan. No setting or agreement removes them.
1.1 Child safety. You must not create, request, share or store:
- child sexual abuse material, whether real, drawn or AI-generated;
- sexual content involving anyone under 18, or characters described as or appearing to be under 18;
- content that sexualises children, or that grooms, exploits or facilitates the abuse of a child; or
- guides or instructions for abusing children or for creating abuse material.
We block this content, close the account immediately and report it to the National Crime Agency or other relevant authorities.
1.2 Intimate images without consent. You must not create sexual or intimate images, video or audio of a real, identifiable person without their consent, including "nudified" or deepfake content.
1.3 Terrorism and violent extremism. You must not plan, promote, fund or support terrorism or violent extremism, recruit for it, or create propaganda for it.
1.4 Weapons capable of mass harm. You must not seek help to develop, produce, acquire or use chemical, biological, radiological or nuclear weapons, or explosives intended to cause harm.
1.5 Serious harm to people. You must not use the Service to plan or incite violence against people, to traffic people, or to encourage or give instructions for suicide or self-harm.
2. Illegal and harmful activity
You must not use the Service to:
- break the law in the UK or the country where you are;
- defraud or deceive people, including scams, phishing, fake reviews, identity theft and impersonating real people or organisations;
- send spam or carry out mass unsolicited messaging;
- attack systems, including creating or spreading malware, gaining unauthorised access, or carrying out denial-of-service attacks, except authorised security testing allowed by section 5.3;
- harass, threaten, bully or defame anyone, or publish someone's private information to harm or intimidate them;
- stir up hatred against people because of protected characteristics such as race, religion, disability, sex or sexual orientation, where that's illegal;
- infringe intellectual property, such as copying protected works or trade marks without permission;
- process personal data unlawfully, including covert surveillance, tracking people without their knowledge, or building facial recognition databases from scraped images;
- trade illegal goods or services, such as illegal drugs, weapons or stolen data;
- run deceptive influence campaigns, such as networks of fake accounts or large-scale disinformation meant to mislead people about elections or public health; or
- make important decisions about people automatically, such as on credit, jobs, housing or access to services, without appropriate human review.
3. Misusing the Service
You must not:
- try to get around our safety checks, keyword filters, rate limits or billing;
- share your API keys, or resell direct access to the Service or to models without our written permission (using the Service inside your own product is allowed);
- create multiple accounts to avoid a suspension, a limit or a ban, or to collect free or promotional credits;
- create accounts automatically or in bulk;
- copy, reverse engineer, scrape or overload the Service, or test its security without our written permission;
- use Outputs to train or improve AI models where the Model Provider's terms don't allow it;
- hide that people are dealing with an AI where the law requires you to tell them; or
- interfere with other users' use of the Service.
4. Adult content
4.1 The Service is only for people aged 18 or over and isn't designed or offered for sexually explicit content. Abliterated models may still produce sexually explicit text if asked. The Service doesn't generate sexually explicit images, audio or video.
4.2 Any sexually explicit content must:
- involve only adults, and fictional characters or consenting adults;
- not depict real, identifiable people without their consent (section 1.2); and
- not be illegal, including extreme pornography under UK law.
4.3 You must not show or share adult content with anyone under 18, or publish it where children can access it.
5. Abliterated models
5.1 Abliterated models have had their tendency to refuse requests reduced. They're intended for lawful work such as:
- authorised security testing and red-teaming;
- research in sensitive fields, such as medicine, law, security or safety;
- creative writing that deals with difficult themes; and
- testing and evaluating AI systems.
5.2 Requests to these models and their responses are logged for 30 days and may be reviewed for illegal content.
5.3 You may only carry out security testing on systems you own, or that you have written permission from the owner to test, within the scope of that permission.
5.4 Everything in sections 1 to 3 still applies. A model being willing to answer doesn't make a request allowed.
5.5 We may remove your access to abliterated models at any time if we think you're misusing them, and they may not be available in some countries.
6. Business customers and their end users
6.1 If you build a product on Kalsenti, you're responsible for how your end users use it. You must:
- have your own terms for end users that are at least as strict as this policy;
- give end users a way to report abuse, and act on reports;
- tell us promptly about any serious misuse you find; and
- where we ask, help us identify which end user was responsible for a request, for example by sending a pseudonymous user identifier with requests.
6.2 If one of your end users breaks this policy, we may ask you to suspend them. If you don't, we may suspend or restrict your account.
7. Model Provider rules
7.1 Some models come with extra usage rules from their Model Provider, shown on the model's page. You must follow them when you use that model.
7.2 If a Model Provider's rules are stricter than this policy, the stricter rule applies to that model.
8. How we enforce this policy
8.1 We use automated safety checks, including image matching against known child sexual abuse material, keyword filtering and random sampling of requests, with human review where needed. When a check blocks a request, we record the reason, not the content. Our Privacy Policy explains how these checks handle your data.
8.2 Depending on how serious a breach is, we may:
- block the request;
- warn you;
- restrict access to certain models or features, such as abliterated models;
- suspend or close your account, in which case unused credits are handled as our Terms of Service explain; and
- report the matter to the police, the National Crime Agency or other authorities, and share information with them where the law requires or allows it.
8.3 We have zero tolerance for child sexual abuse material. Any account involved is closed immediately and reported, with no warning.
9. Reporting abuse
If you see the Service being misused, or find a security problem, email support@kalsenti.ai with "Abuse" or "Security" in the subject line. Tell us what happened, when, and any request IDs or links. Don't send us child sexual abuse material: describe it and report it to the Internet Watch Foundation at iwf.org.uk.
10. Appeals
If you think we've restricted or closed your account by mistake, email support@kalsenti.ai within 30 days. A person will review the decision and reply. We won't restore an account where we've confirmed child sexual abuse material.
11. Changes
We may update this policy as the law, our Model Providers' rules or the Service change. We'll post the new version here with a new date, and tell you by email about significant changes.