Acceptable use policy

This policy is short because it is meant to be read. It applies to everything you train, serve or generate using the service, and it applies in addition to the acceptable-use terms of whichever base model you fine-tune — those travel with the model and we cannot waive them.

You may not use the service to

• ·Produce material that sexually exploits children, or that depicts real people in sexual content without their consent.
• ·Generate content designed to harass, defame or impersonate a specific person, or to fabricate their words or actions in a way intended to be believed.
• ·Build or operate systems for large-scale deceptive influence, spam, or fraud — including generating fake reviews, fake identities or synthetic engagement.
• ·Develop weapons, or provide meaningful uplift toward chemical, biological, radiological or nuclear harm.
• ·Attack computer systems you are not authorised to test, or develop malware intended for unauthorised use.
• ·Make consequential decisions about individuals — credit, employment, housing, insurance, benefits, criminal justice — without human review and without meeting the legal obligations that attach to those decisions.
• ·Train on data you do not have the right to use, including personal data processed without a lawful basis.
• ·Circumvent metering, quotas or spend limits, or resell raw compute access.

Things that are explicitly fine

Security research and testing against systems you own or are authorised to test. Content moderation and safety classifiers, which necessarily process material they exist to detect. Red-teaming your own models. Medical, legal and financial applications, provided a qualified human remains responsible for the decision.

What happens if you breach it

We will contact you where it is safe and practical to do so, and give you the chance to fix it. For serious breaches — the first three items above in particular — we will suspend immediately and, where the law requires it, report. We do not monitor the content of your datasets or your inference traffic as a matter of course; we act on reports and on evidence.

Reporting

If you believe something on this platform breaches this policy, write to hello@onrup.com with enough detail for us to investigate. Security vulnerabilities go to security@onrup.com.

Onrup may update this policy as new categories of misuse appear. Material changes will be announced before they take effect.