PRIVACY & Terms

We log requests. Never your users.

Foreshock’s privacy posture is the product, not a footnote. The short version: we collect geometry and scores, never content, and we never build a profile of anyone.

Privacy policy

We never collect data about your users.

If you submit text to be classified, all we log is the time of the request, which model was used, how long it took, and whether anything went wrong. As soon as the specimen text is converted to vectors, we have no further practical use for it. We run no static analysis beyond checking that it is not too large for the embedder’s context window. Per our terms, a user’s computer can never initiate a request to us directly: it has to come from you, because it must carry your API key.

We never share your data with third parties.

The only exceptions are capturing payment and delivering services you requested. For example, if you order on-prem equipment from us, expect us to share your shipping information with a logistics company such as FedEx, UPS or USPS. We never share information purely for marketing purposes without your express, informed consent.

We do not use third-party AI.

Foreshock does not use third-party AI for classification or verification of any kind. Our two infrastructure providers are OVHCloud, from whom we buy compute, and Modal, from whom we buy GPU. The code and the inference belong to us. We are building our own custom LLM for verification escalation, and it will also be in-house, trained and tuned by us.

Logged
Request metadata

Time, model used, duration, and error state. That is the whole log.

Discarded
The text

Converted to vectors, then dropped. No static analysis beyond a size check.

Never
PII or profiles

No user data, no regional data, no marketing without informed consent.

Terms of service

Requests carry your key.

You may never have a user’s computer initiate a classification request to us directly. Every request must originate from you and must contain your API key. This is what makes the no-PII guarantee possible: we only ever see the specimen text and your key.

Open corpus, open license.

Aldous is open source. Splinter and Libsplinter are Apache2 licensed. You can inspect and edit the phrases behind every sensor and shunt. There is no such thing as real security through obscurity, so we do not pretend otherwise.

Fair use and rate limits.

RSS ingestion guards against temporal misattribution and rate-limits against runaway apps. API throughput is governed by your tier. Abuse that threatens shared infrastructure may be throttled to protect every other community on the platform.

This page summarizes intent for the website. Your binding agreement is the signed contract or click-through terms presented at signup.

Questions about data

Ask us anything about how this works.

Contact us Read the method