Acceptable Use Policy

Effective date: [EFFECTIVE DATE] · Version: 1.0-draft · Part of the FrugalAI Terms of Service.

You are responsible for your use of FrugalAI, for your prompts, applications, and end users, and for the legality of the content you submit and the outputs you use. Because FrugalAI stores prompt hashes rather than content by default, we cannot and do not review your content; that responsibility is yours.

Prohibited uses

Do not use the Service to:

1. Violate any law, or infringe or misappropriate anyone's rights, including privacy and intellectual property rights.
2. Create, request, or distribute child sexual abuse material of any kind, including AI-generated. We report apparent CSAM to the National Center for Missing and Exploited Children and law enforcement as required by law.
3. Generate or distribute malware; conduct or facilitate phishing, fraud, or scams; or plan or facilitate violence, terrorism, or the acquisition of weapons capable of mass harm.
4. Send spam or unlawful communications, or violate marketing and robocall laws.
5. Harass, threaten, or abuse people, or generate content that sexualizes minors or non-consenting persons.
6. Violate the usage policies of the model providers you route to. Your provider agreements travel with your traffic.
7. Probe, scan, or test the vulnerability of the Service, attempt to access other customers' data or defeat tenant isolation, or interfere with the Service's operation, except through the coordinated disclosure process at frugalai.io/security. Do not circumvent rate limits, plan limits, or budget enforcement.
8. Attempt to defeat, manipulate, or poison the quality-evaluation or routing systems, including via prompt injection against the judge.
9. Resell or white-label the Service without a written agreement with us.
10. Misrepresent AI-generated output as human-authored where the law requires disclosure, or use the Service to make fully automated decisions with legal or similarly significant effects on individuals without the human review the law requires.

High-stakes uses

If you use model outputs in domains where errors can cause serious harm (medical, legal, financial advice, employment, credit, insurance, safety-critical systems), you must keep a qualified human in the loop and comply with the laws of that domain. FrugalAI routes and meters traffic; it does not validate outputs for these uses.

Enforcement

We may investigate suspected violations using the metadata available to us, and may suspend or terminate per the Terms of Service. For dangerous or unlawful activity we may suspend first and notify after. We may report unlawful activity to authorities where required. Report abuse to abuse@frugalai.io.

Benchmarks

You may publish benchmarks of the Service if they are accurate, current, and include your methodology.