Acceptable Use Policy

This policy applies to everything reached through SoreQen: the chat service, the API, and anything you build on top of either. It forms part of the Terms of Service, and it binds your end users as well as you — if you build a product on the API, you are responsible for what people do with it.

• ·Sexualise children. Generating, soliciting or handling sexual content involving minors is prohibited absolutely and is reported to the authorities.
• ·Build weapons or plan violence. This includes chemical, biological, radiological, nuclear and high-yield explosive weapons, and planning or inciting an attack on people or infrastructure.
• ·Attack computer systems. Malware, ransomware, credential harvesting, denial-of-service tooling, or scaled exploitation of systems you do not own or have written authorisation to test.
• ·Deceive people at scale. Disinformation campaigns, coordinated inauthentic behaviour, impersonating a real person or organisation, or generating fraudulent documents, reviews or identities.
• ·Run surveillance or profiling that tracks individuals without consent, infers protected characteristics to discriminate, or supports biometric identification of people in public.
• ·Give unqualified professional advice. Do not present output as legal, medical, financial or psychological advice from a qualified professional, and do not use it to make automated decisions about credit, employment, housing, insurance or benefits without human review.
• ·Harass or exploit. Targeted abuse, threats, stalking, doxxing, or non-consensual intimate imagery of anyone.
• ·Break the law where you or your users are, in any other way.

If you deploy SoreQen in a product that talks to people, they must be able to find out that the replies are generated. You need not label every message, but you must not design the product to make a person believe they are speaking to a human when they are not.

• ·Do not attempt to extract, copy or reconstruct model weights.
• ·Do not use output to train, fine-tune or distil a competing model.
• ·Do not circumvent rate limits, quotas, spend caps or safety measures, including by creating multiple accounts to evade a limit.
• ·Do not resell raw API access as though it were your own model service. Building a product on the API is fine; reselling the endpoint is not.
• ·Do not share an API key outside your organisation, or embed one in client-side code where users can read it.

The distinction throughout is between discussing a subject and doing harm with the output. The models will engage with violence, extremism, drugs, sexuality, self-harm, crime and other difficult subjects in the ordinary course of research, journalism, education, fiction, medicine, law and security work.

Security professionals may use the models for authorised testing, vulnerability research and defensive work. What is prohibited is unauthorised attack, not the subject matter.

We monitor for abuse using aggregate signals — request patterns, volumes, error rates — rather than by reading conversations. Where a signal warrants it, a narrow and logged review may follow.

Depending on severity we may warn you, rate-limit an account, revoke a key, suspend access, or close the account. For the most serious categories, particularly child sexual abuse material and credible threats to life, we act immediately and report to the relevant authorities.

Where circumstances allow, we will tell you what was wrong and give you a chance to fix it. Where they do not, we will act first and explain afterwards.

Report misuse to abuse@zorqelisai.com. Report a security vulnerability in the service itself to security@zorqelisai.com — we will not pursue good-faith research that respects user privacy and gives us reasonable time to fix the issue before disclosure.

This policy will change as the models and their uses change. Material changes are announced at least 30 days in advance, except where a change is needed immediately to prevent harm.