AI review of legal documents
Contracts that get read without consequences.
Drop in a contract and we strip out the risk layer: ambiguous terms, void clauses, liability imbalances, payment and force-majeure traps. Personal data and trade secrets stay inside the perimeter — only de-identified text reaches the model.
Text is de-identified before it reaches the model.
Review · contract no. 12/2026
risks found: 3Draft agreement
The Supplier — ORG_1, VAT INN_1, and the Buyer — ORG_2, the Parties…
4.2For late payment the Buyer pays a penalty of 0.5% per day , while the Supplier pays 0.01% for late delivery.
5.1The Parties are liable in full for any non-performance.
6.1The warranty period is agreed separately .
Comments
Liability imbalancecl. 4.2
Buyer penalty 0.5% per day vs 0.01% for the supplier.
High
No force majeurecl. 5
No clause on events beyond the parties’ control.
Medium
Warranty period undefinedcl. 6.1
Fix the term and scope of the warranty.
Low
- 3.4 s
- per page reviewed
- 40+
- risk types
- 0
- sensitive data in the model
- GDPR · 152-FZ
- jurisdictions
How it works
Three steps — from text to risk cards.
Step 1
Upload
Paste a contract or upload a file (docx / pdf / scan).
Supply agreement no. 12/2026.docxStep 2
De-identify
INN, OGRN, accounts, VAT, IBAN, names and trade secrets become placeholders.
Acme Ltd, VAT 7701234567
<<ORG_1>>, VAT <<INN_1>>Step 3
Review risks
The model returns cards: severity, clause, recommendation and grounding.
High · cl. 5.2 · penalty 0.5% per day
What’s inside
No “magical” promises — only what actually works.
- Contract audit
- Structured report per risk: severity, clause, legal grounding and a recommendation.
- Analysis from your side
- Pick your side of the deal: risks and recommendations are assessed from your position, and clauses that work against you are highlighted in the report.
- Legal citation checks
- Every cited article is checked against a statutory corpus. The report shows which references are confirmed.
- Protocol of disagreements
- Auto-generated amendments to terms with legal rationale.
- De-identification
- INN, OGRN, accounts, VAT, IBAN, names, emails become placeholders before reaching the LLM.
- Regions & compliance
- EU (GDPR) and RF (152-FZ): separate providers, payments and data residency.
- AI credits
- A transparent append-only ledger — purchase and spend per actual token.
- Formats & export
- Supports docx / pdf / scans; export the audit report to DOCX.
Security
Data never leaves the trusted perimeter
Only de-identified text reaches the LLM. Re-identification happens on our side — you see real values, the model never does.
- De-identification
- INN, OGRN, accounts, VAT, IBAN, names → placeholders before the model.
- Data residency
- Pick a region — EU or RF: its own LLM providers and payments.
- Operation audit
- An idempotent ledger — every credit spend and top-up is visible.
Sample report
This is what the final report looks like: an audit of an anonymised standard supply contract — risks, legal grounds and a redline protocol.
Demonstration document with no client data.
Frequently asked questions
Does the service replace a lawyer?
No. The result is a machine analysis for information purposes: it helps you find questionable clauses quickly, but it is not legal advice. The decision to sign is yours or your lawyer’s.
What exactly is sent to the model?
Only de-identified text. Tax IDs, registration numbers, accounts, names and other details are replaced with placeholders before sending. Restoration happens on our side, so the model never sees the real values.
Which files can I upload?
Text, docx and pdf. Scans and spreadsheets (jpg, png, xlsx) are accepted too, but work in an experimental mode: the extracted text may contain noise and errors.
How is the price of a review calculated?
The price depends on the volume of text. Before you start the analysis you see the cost in credits and your balance, so there are no surprise charges.
Is there a subscription?
No. You buy a credit package for the period you need. There is no auto-renewal, and unused credits expire when the period ends.
Where is data processed?
You choose a region: EU (GDPR) or RU (152-FZ). Each region uses its own providers, payment systems and data storage.
Start with a single contract.
Your first review takes a few clicks.