How Recensa handles your documents and AI

Last updated: June 29, 2026

Recensa reviews confidential documents — contracts, agreements, financial and legal filings. We know the first question any serious buyer asks about an AI tool is simple: where does my document go, and could it ever be used to train someone's model?

This page answers that plainly.

The short version

• Your documents are never used to train any AI model — not ours, and not any of the third-party models we send them to.
• Documents are processed to produce your review, then handled under a strict retention and deletion policy.
• Our public verification record stores a cryptographic hash and job metadata only — never the contents of your document.

How a document moves through Recensa

When you submit a document for review, its text is analyzed by a third-party AI model operating under Recensa's commercial agreements with that provider. Anthropic is the primary reviewer. OpenAI and Google are configured as backups and receive document text only when the primary provider is unavailable. We name all three because you have a right to know exactly who can process your data.

Each of these providers, on the commercial API terms Recensa operates under, contractually commits that data sent through their API is not used to train their models. This is a property of the commercial agreement itself — not a setting we have to remember to switch off.

Retention and deletion

• At the AI providers: under their commercial terms, document text may be retained briefly (typically up to 30 days) solely for security and abuse monitoring, after which it is deleted. It is never used for training during or after that window.
• At Recensa: when a job completes, your uploaded documents and generated outputs are scheduled for automatic deletion — by default about 7 days after completion, and about 24 hours when you enable sensitive-style handling. You can delete an individual job yourself at any time. You can also request deletion by emailing privacy@recensa.ai.

The Recensa Checked seal

When Recensa issues a Checked seal, the public record behind it contains a cryptographic hash of the document and basic job metadata — not the document itself. Anyone verifying a sealed document is checking a fingerprint, not reading your content. The seal confirms the file is unchanged since the check; it does not certify that the content is correct. Correctness of any edit always remains in your hands through the review-and-accept flow.

What we keep private about how Recensa works

We are transparent about who processes your data and how it is protected. We do not publish the internal design of our review engine — how the panel is orchestrated and how results are produced. That protects the integrity and independence of the assurance layer. Transparency about data handling and secrecy about engine internals are not in tension: you can know exactly where your document goes without us exposing how the verification is built.

Questions

Email security@recensa.ai for security and data-handling questions, or privacy@recensa.ai for privacy, retention, and deletion requests.

---

Recensa's data-handling commitments reflect the commercial API terms of our AI providers as of the date above. Provider terms are summarized in good faith; the controlling terms are each provider's own commercial agreement and Recensa's contract with them.