OneHot by Sqwish Labs · Legal

Acceptable Use Policy

In force from 1 October 2026

This policy says what you may not do with OneHot, and what you must do when you use it in decisions about people. It is part of our Terms of Service. It applies to you, to everyone who uses your account, and to anyone who uses OneHot through a product you build (your end users). You are responsible for making sure they follow it.

OneHot is a classifier, so many good uses involve harmful material: spotting abuse, scams, jailbreaks or unsafe requests so they can be stopped. That is allowed. What this policy forbids is using OneHot to cause harm, not to detect it.

On this page

  1. The law and other people
  2. Decisions about people
  3. The data you send
  4. Protecting the service
  5. Referrals and promotions
  6. Sanctions and export controls
  7. If this policy is broken
  8. Changes to this policy

1. The law and other people

Don't use OneHot:

  • for anything illegal, or to help anyone break the law, including laws on data protection, privacy, consumer protection, equality, intellectual property and electronic marketing;
  • to threaten, harass, bully, defame or abuse anyone, or to promote violence, terrorism or hatred against people for who they are;
  • to defraud or deceive people, for example with scams, phishing, impersonation or fake reviews;
  • to develop, make or use weapons that can cause mass casualties, including chemical, biological, radiological and nuclear weapons;
  • to create or spread malware, or to attack, break into or disrupt any system or network you don't have permission to test;
  • to send spam or unsolicited marketing;
  • to infringe anyone's intellectual property or other rights.

Never send us child sexual abuse material, and never use OneHot in any way that sexualises or endangers children. If we become aware of such material, we report it to the authorities. If you build tools that protect children and need to handle other harmful material about them, such as grooming messages, talk to us first.

2. Decisions about people

Some uses are never allowed. Don't use OneHot:

  • to score people on their social behaviour or personal traits in a way that leads to unfair or unjustified treatment;
  • to manipulate people, or to exploit their age, disability or social or economic situation, in ways likely to cause them significant harm;
  • to infer what someone is feeling at work or in education, except for medical or safety reasons;
  • to infer someone's race, political opinions, trade union membership, religious or philosophical beliefs, sex life or sexual orientation from biometric data;
  • to predict whether a person will commit a crime based only on profiling or their personality traits;
  • to build or expand facial recognition databases from images scraped from the internet or from CCTV;
  • to monitor, track or profile people covertly, or without a lawful basis;
  • to discriminate unlawfully against anyone.

Other uses are allowed only with safeguards. If OneHot's answers help decide something with legal or similarly significant effects on a person, such as access to work, credit, insurance, housing, education, healthcare, benefits, essential services or legal status, you must:

  • have a person with the authority and competence to change the outcome review it before it takes effect, unless the law where you operate allows the decision to be fully automated and you give people the safeguards it requires, such as a way to contest the outcome and ask for a person to review it;
  • tell people that automated tools are involved, and how to challenge the outcome, wherever the law requires it;
  • test the model on your own data for accuracy and for unfair differences between groups before you rely on it, and keep checking it;
  • meet the laws on automated decision-making and artificial intelligence that apply where you and the people affected are, such as UK and EU data protection law and the EU AI Act.

Don't use OneHot as the only basis for a medical diagnosis or treatment, for legal advice, or in systems where a wrong answer could injure or kill someone, such as controlling vehicles, machinery, medical devices, critical infrastructure or weapons.

3. The data you send

  • Only send content you have the right to send. Where it includes personal data, you must have a lawful basis and must have told the people concerned whatever the law requires.
  • Send health data, data revealing ethnicity, religion, sexual orientation or political views, criminal records, children's data or government identifiers only where you need to, the law allows it, and you have done what the law requires first, such as a data protection impact assessment.
  • Never put payment card numbers, passwords, API keys or other secrets in decision content or datasets.
  • Don't send health information that US law (HIPAA) protects, unless we have signed a business associate agreement with you. We don't offer one yet.

4. Protecting the service

Don't:

  • try to reverse engineer, decompile or extract our models, their weights, our prompts, our training data or our source code, except where the law allows you to despite this policy;
  • use OneHot, or what it returns, to train or fine-tune a model outside OneHot that competes with it, or to build a competing service. Training models inside OneHot for your own account, using answers in your own products, and comparing OneHot's results with other models' results are fine;
  • get around rate limits, usage limits, credit checks, the waitlist, access controls or security measures;
  • create more than one account to get extra free credit, to avoid limits, or to come back after we suspended or closed an account;
  • share API keys outside your organisation, or let people outside it use your account;
  • resell, sublicense or offer raw access to OneHot as a service of its own unless we agree in writing. Building your own product on top of OneHot is fine;
  • scan, probe or load-test OneHot, or test its security, without our written permission;
  • overload the service, interfere with other customers, or send requests designed to make it fail.

If you come across a security weakness, tell us at [email protected] and give us a reasonable time to fix it before you tell anyone else. We won't take action against you for finding and reporting it in good faith, as long as you looked at no more data than you needed to show the problem and didn't harm the service or other customers. This applies to OneHot only, not to our providers' systems, and it doesn't cover denial-of-service testing, social engineering or looking at other customers' data.

You may publish your own tests of OneHot. Describe your method and data fairly, so others can check your results.

5. Referrals and promotions

  • Share your referral link only with people who would welcome it. Don't spam, don't post it where that isn't allowed, and don't bid on our name in search ads.
  • Don't refer yourself, your own other addresses, or accounts you control, and don't create accounts to collect rewards or starter credit, or to move up the waitlist.
  • Describe OneHot and the offer accurately, and never suggest that you speak for us.

We may withhold or remove credit obtained in breach of this section, as the Terms of Service describe.

6. Sanctions and export controls

Don't use OneHot, or let anyone use it, in or for the benefit of a country or region under comprehensive UK, EU, US or UN sanctions, or for anyone named on a sanctions list in those places. Don't use it for an end use that export control laws forbid, such as developing weapons.

7. If this policy is broken

We may investigate suspected breaches, and ask you to explain or stop. Depending on how serious the breach is, we may block requests, suspend or close the account, withhold or remove credit, and report the matter to the authorities. We will usually warn you first and give you a chance to put things right, unless the harm is serious or urgent, the law stops us, or warning you would help someone evade detection.

To report misuse of OneHot, write to [email protected].

8. Changes to this policy

We may update this policy as the law, the service and the risks change. We'll email account holders at least 30 days before a material change takes effect, unless the law or a new risk to people or the service needs it to apply sooner. The date at the top shows when this version took effect.

Legal

Overview Terms of Service Acceptable Use Policy Privacy Policy Data Processing Addendum Subprocessors