Acceptable use

PrinciplesCopy Anchor Link

Sieve models are decision models. They return probabilities. Use them to classify, filter, route and rate content, with a person accountable for decisions that matter.

You must notCopy Anchor Link

• Use the API for anything illegal, or to create or distribute child sexual abuse material. We preserve and report such material as the law requires.
• Probe, overload or evade the rate limits or security of the service, or share keys.
• Resell access in a way that breaks the Master Customer Agreement or your Order.
• Use outputs as the sole basis for decisions that need human judgement where the model card says it is not suitable. Read the limits on each model card.

Model termsCopy Anchor Link

Some models are built on third-party base models, and their terms pass to you, whether you use the model through the API, through Vault or as downloaded weights. Sieve-31B is built on a Gemma model, so Gemma's terms and prohibited-use policy apply. Each card lists its base model and licence.

Downloaded open weights are governed by their own licence, as listed on the model card. This policy covers your use of the hosted Services.

EnforcementCopy Anchor Link

We review content flagged by automated safety checks, warn, throttle or suspend accounts, and offer an appeal. Flagged content handling follows Data retention. To appeal a decision, write to support@sthanika.ai.

Next stepsCopy Anchor Link