Responsible AI

AI that drafts. Professionals who decide.

Contract review affects people's money, jobs and businesses. These are the commitments behind how we build TermOwl, and how you can check that we keep them.

A professional stays in control

TermOwl is built for lawyers, in-house counsel and paralegals working under a lawyer's supervision. Before every analysis the user confirms that a qualified professional will review the output before it is relied on or shared. We do not market TermOwl to consumers as a substitute for legal advice.

Clear disclosure

Every session says plainly that results are generated by AI (Claude, by Anthropic), are drafts, can contain mistakes and are not legal advice. TermOwl is for users aged 18 and over.

Every claim is checkable

Each finding quotes the exact clause it relies on, and we verify every quote against the uploaded document. Legal citations are verified against the official text of the cited article, with a link to the source. Anything that cannot be verified is marked for checking.

Official sources, not memory

When TermOwl applies Ukrainian law, it may cite only articles from the official Civil Code text we provide, with the edition stated. It also detects references to law that is no longer in force.

Measured and published

We test TermOwl against contracts with hidden problems, clean controls and manipulation attempts, publish the results including what went wrong, and run the tests again before changes ship.

Resistant to manipulation

Uploaded documents are treated as untrusted content: instructions inside them are ignored. In our benchmark, a hidden “report low risk” instruction was ignored in every run.

Privacy by design

Documents are processed in memory and never stored. Analyses run through the Claude API, which does not use API data to train models. Saved reports are kept only when a signed-in user chooses to save them, in the EU.

Usage policy and accountability

TermOwl is built on the Claude API and follows Anthropic's Usage Policy for legal use cases: a qualified professional reviews AI output before it reaches a client, and users are told when they are interacting with AI. If you see output that is wrong, biased or unsafe, tell us at hello@termowl.com; we read every report.