Zero Data Retention

Spice Finance Inc, doing business as Pareto Inference ("Pareto Inference," "we," "us," or "our"), does not intentionally keep the prompts, messages, files, or other inputs you send to our inference API, or the outputs it returns (together, "Customer Content"), in persistent storage once a request has finished.

While a request is being served, Customer Content may be held briefly in memory, including in prompt caches, only as long as reasonably needed to run inference and operate the Service efficiently.

We do not use Customer Content to train or fine-tune AI models unless you separately and expressly opt in. Using the Service or accepting our Terms of Service is not an opt-in.

We keep operational records about requests, but not their content. These may include:

• account, organization, and user identifiers;
• request identifiers and API-key identifiers;
• the model requested and the backend that served it;
• token counts, cache usage, prices, and charges;
• timestamps, latency, status codes, and error types;
• IP addresses and security-event records;
• billing, account, and technical information.

We use these records to operate, secure, support, bill for, analyze, and improve the Service. They do not include the text of prompts or outputs.

Most requests are served on inference hardware we operate. When a request cannot be served there, we instruct OpenRouter to use only providers it designates as zero data retention. Those providers process Customer Content under their own terms and technical controls.

This policy covers Customer Content sent through our standard inference API. It does not cover information you give us in other ways, such as account details, billing information, support requests, sales conversations, feedback, or other correspondence. Our Privacy Policy covers that information.

This policy also does not cover data kept by systems you control, including your own applications, logs, and third-party tools or integrations.

If a signed agreement, order form, data processing addendum, or configuration we have agreed with you in writing conflicts with this policy, that agreement controls.

We may preserve or disclose information when the law requires it, or when reasonably necessary to investigate fraud or unauthorized access, enforce our terms, or protect Pareto Inference, our customers, or others. This does not require us to keep Customer Content we would otherwise not keep.

Customer Content you ask us to store, content you send us for support or debugging, and content you opt in to sharing for evaluation or training is handled under the written terms that apply to it.

You are responsible for the Customer Content you send and for how your own applications, logs, and integrations handle it. Do not send information that our Terms of Service or the applicable model's policies prohibit. You are responsible for deciding whether the Service meets your legal, regulatory, and contractual requirements.

This policy supplements our Terms of Service and Privacy Policy. It describes our current practice. It does not replace a signed agreement or create a service-level commitment.

We may update this policy as the Service, our practices, or legal requirements change. We will post the updated version and revise its effective date.

Questions about this policy may be sent to admin@paretoinference.com.