Privacy Policy

What We Collect

• TOS Text: The Terms of Service text you submit for analysis. This is processed by our AI and temporarily stored to generate your report.
• Company Names: If you choose to share your analysis with the community library, the detected company name is stored publicly.
• View Counts: We track anonymous view counts on public analyses to show popularity.

Data Retention

• Private Analyses: Analyses not shared to the library are stored for 30 days, then automatically deleted.
• Cached Results: We use Redis caching to improve performance. Cached data is temporary and automatically expires.
• Public Library: Analyses shared to the community library are stored indefinitely to benefit other users.

Third-Party Services

To analyze your TOS documents, we send the text to a gateway we run ourselves, on the same server as this app. The gateway doesn't analyze anything itself — it forwards your document to one of four third-party AI providers and returns the result. The gateway sends your document to one provider at a time, starting with Cerebras; if that provider is unavailable, errors out, or returns a suspiciously empty result, the gateway retries the same document with the next provider in line — so more than one provider can end up seeing the same document within a single analysis. Which provider (or providers) that ends up being depends on live availability, not on anything you pick, and we can't tell you in advance.

What each provider is allowed to do with your document once it arrives is not the same, and it matters because you don't get to choose which one you land on:

• Cerebras (tried first, via two different models before the gateway moves on): Cerebras's Inference Terms and Conditions limit its use of submitted content to providing the service — “we may access and use the User Content to provide the Services and comply with applicable law” — and separately exclude submitted content from the data it's allowed to use to improve its products. Cerebras does not use your document to train its models. Cerebras Privacy Policy
• Cohere (tried third): Cohere's Privacy Policy states that for Trial API users — the free access tier this gateway uses — Cohere “may use this input and output data to conduct research and development,” including to “improve the performance and safety of Cohere Products.” The opt-out for this exists only for enterprise customers with a credit card on file, which doesn't apply here. Cohere may use your document to improve its models, and we have no way to opt out of that on this tier. Cohere Privacy Policy
• SambaNova (tried fourth): SambaNova's Cloud End User License Agreement limits its rights over submitted content to what's “necessary to provide the Service to you or as may be required by law, and for no other purposes,” and separately excludes submitted content from the usage data it can apply to improve its products. SambaNova does not use your document to train its models. SambaNova Cloud EULA
• NVIDIA (last resort, used only if the other three providers are all unavailable): NVIDIA's API Trial Terms of Service — which govern this kind of free access — state that NVIDIA collects “User Content and Generated Content to improve NVIDIA products and services, including AI models.” NVIDIA does use your document to improve its models, including training. NVIDIA API Trial Terms of Service

In short: two of the four providers in this cascade (Cerebras, SambaNova) contractually limit themselves to processing your document for your request and nothing else. Two (Cohere, NVIDIA) may use it to improve their own models. Because routing is automatic, we can't tell you after the fact which one(s) touched any specific analysis.

Separately, our own gateway keeps a call log of metadata only — which model and provider handled the request, how long it took, and how many tokens were used. It does not log the text of your document. The gateway also has an optional feature to store full conversation history; we do not enable it for TOS analysis.

Community Library

• Public by Choice: Sharing to the library is opt-in. Each analysis is private by default — you must explicitly toggle “Share with Community” on before clicking Analyze.
• Permanent Records: Once shared, analyses become part of our public database to help other users research companies.
• No Personal Data: Public analyses contain only the company name, risk assessment, and clause analysis—no information about who submitted it.