The deal waits for the NDA
A prospect wants to share requirements, a supplier wants to see drawings. Nothing moves until someone in legal has read the confidentiality agreement.
AI NDA review for legal, sales and procurement
NDA review AI that knows your fallback positions — so legal can focus on the NDAs that deviate.
Forward an NDA to a dedicated inbox and Treasure Contracts checks it against your NDA playbook, criterion by criterion, with a compliant wording suggestion for every finding.
Mutual or one-way, term, carve-outs, residuals, non-solicit, governing law: your rules are applied the same way to every confidentiality agreement, including the version that comes back from the other side.
ISO 27001 · Hosted in Frankfurt · GDPR · 4-week right of withdrawal
In short
How does NDA review with AI work?
NDA review AI compares each clause of a non-disclosure agreement with your company's standards. In Treasure Contracts, an NDA arrives by upload or by email to a plus address, runs against your NDA playbook and receives a traffic light per criterion, such as term, carve-outs or residuals. Each finding comes with compliant wording, and every returned version is reviewed again automatically.
Key facts
- Plus addresses route incoming NDAs straight to the NDA playbook, protected by DMARC/SPF checks and a sender whitelist
- Result per criterion as compliant, moderate or critical, with a suggested compliant wording
- Every new version is stored immutably, compared as a redline and reviewed again
The problem
NDAs are low risk per document and high cost in total.
No single NDA is worth an hour of legal time. But dozens a month, each read in full, add up to a bottleneck that slows down sales talks, supplier onboarding and partnership projects.
Counterparty paper, every time
Your own template is rarely used. Each NDA is drafted differently, so the one-way clause, the residuals wording or the unusual term has to be found anew.
Small edits in the second round
The signed-off draft comes back with a changed definition of confidential information or a new non-solicit clause. Few people compare it line by line.
Rules exist, but only in heads
Which carve-outs are acceptable, how long confidentiality should survive, which governing law is fine: this knowledge sits with two people and is applied unevenly.
Your outcome
Standard NDAs pre-reviewed against your standard. Deviations go to legal with a proposed fix.
The goal is not to stop reading NDAs. It is to stop reading the ones that already match your standard.
Your outcome
The business forwards an NDA and gets back a traffic light per criterion. Green NDAs move on, amber ones come with wording to send back, red ones go to legal.
Why it works
Your legal team defines and confirms every criterion. Golden-set tests with synthetic NDAs show whether the playbook flags what it should before it goes live.
How fast
In a 30-minute demo we run one of your real NDAs. In the 10-day Proof Sprint, your NDA process runs live with risk rating, redlines and audit trail.
How little effort
We derive the playbook from 10–20 of your historical NDAs and existing guidelines. Afterwards, colleagues simply forward NDAs to an email address.
Features
What NDA review in Treasure Contracts includes
Everything listed here is live in the product today.
NDA playbook with your criteria
Define criteria, targets and no-gos, for example: mutual obligations only, confidentiality term of at least three years, standard carve-outs present, no residuals clause, governing law German. Your party perspective is part of the playbook.
Traffic light and compliant wording
Each criterion is rated compliant, moderate or critical with a short rationale. For every finding, Treasure proposes wording that fits your playbook, ready to send back.
NDA intake by email
A plus address such as an NDA review inbox routes forwarded emails and attachments directly to the NDA playbook. DMARC/SPF checks and a sender whitelist per tenant keep the inbox clean.
Re-check on returned versions
When the counterparty sends a new draft, it becomes a new immutable version. Treasure shows a side-by-side redline and reruns the review, so a quietly narrowed carve-out is flagged.
Deviations from your template
Beyond the criteria, the review flags general risks, contradictions within the NDA and deviations from your reference NDA, which helps when the other side uses its own paper.
Human in the loop
Reviewers add risks the AI missed and give feedback on each result. Approval steps with a defined approver (Approval, €190/month, included from Business) and the audit trail document who cleared which NDA.
How it works
From inbox to cleared NDA in four steps
Send one of your NDAs
In a 30-minute demo with the founders, we review a real NDA from your inbox against example criteria. No slides.
Write down your NDA standard
Together we turn your guidelines and 10–20 past NDAs into criteria, targets and no-gos. Legal confirms each rule.
Test with synthetic NDAs
Golden-set tests check whether the playbook catches one-way clauses, missing carve-outs or long non-solicits before live use.
Switch on the email route
The team forwards NDAs to the plus address. Results appear per criterion, and every returned version is checked again.
Comparison
Reviewing NDAs manually, with a general chatbot or with Treasure Contracts
What each approach means for turnaround, consistency and documentation.
| Criterion | Manual review | General AI chat | Treasure Contracts |
|---|---|---|---|
| Intake | Email to a lawyer, then wait | Copy and paste into a chat | Forward to a plus address routed to the NDA playbook |
| Standard applied | Individual experience | Depends on the prompt | Your NDA playbook, confirmed by legal |
| Output | Comments in the document | Free text | Traffic light per criterion plus compliant wording |
| Returned drafts | Compared by eye | Pasted again | New version, redline and automatic re-review |
| Record | Scattered emails | Chat history; no per-contract audit trail like in Treasure | Audit trail with version, result and feedback |
Data handling of general AI assistants depends on the provider, plan and settings.
In detail
NDA review AI in practice: which criteria belong in the playbook
Mutuality and the definition of confidential information
The first two questions in almost every NDA playbook are whether obligations run both ways and how broadly confidential information is defined. A one-way NDA can be fine when you only receive information, but not when your engineers will share designs. A definition that covers everything disclosed in any form, marked or not, may be unworkable for the receiving party. Treasure Contracts lets you express both as criteria with your party perspective, so the same clause can be green in one context and amber in another.
Term, carve-outs and residuals
Confidentiality periods vary widely, and many companies accept a range rather than one fixed number. Standard carve-outs cover information that is public, already known, independently developed or received from a third party; their absence is a classic red flag. Residuals clauses, which allow use of information retained in memory, are often a no-go. Each of these can be set as a target or a no-go in the playbook, and the review states for each one whether the NDA meets it and what compliant wording would look like.
Non-solicit and governing law
NDAs increasingly carry clauses that are not about confidentiality at all, such as non-solicitation of employees or customers, or a contractual penalty. These are easy to miss when a reviewer is scanning for the usual points. A playbook criterion checks for them in every review. The same applies to governing law and jurisdiction: if your standard is German law with a court in Munich, a foreign law clause is flagged in the review rather than discovered after signing.
Why the second draft matters more than the first
Most NDA risk enters during negotiation. The other side accepts nine of your ten comments and slightly rewrites the tenth. Treasure stores every draft as an immutable version, shows a side-by-side redline and reruns the NDA review on each new version. Combined with email intake through a plus address, this turns NDA handling into a routine flow: forward, read the traffic light, send back the suggested wording, and escalate only what is still red.
Your first step
Prove it first. Decide second.
You don't have to take our word for it. You see Treasure on your own contracts — first in a free 30-minute session, then in a 10-day Proof Sprint with measurable results.
For NDA review, that means: we turn your NDA standard into a playbook, connect a plus address for incoming NDAs and compare the traffic lights with your lawyers' judgement on 10–20 real NDAs.
1
Start free: your contract, reviewed live
30 minutes with the founding team. Bring a real contract and we show Treasure live on your clauses. No slides, 24-hour response guarantee — including a free AI Readiness Assessment.
Pick a time2
What you get in the 10-day Proof Sprint
- Baseline for cycle time, volume and legal effort
- Testable rules derived from your existing contracts and decisions
- Live test on 10–20 historical contracts with risk rating, redlines and audit trail
- Sign-off of the rules and jointly defined success criteria
- ROI analysis and a concrete 30-day plan for your first live process
€4,900 net, one-off · fully credited towards implementation if you sign within 30 days
Results guarantee: If Treasure misses the acceptance criteria agreed in writing up front, we keep working for up to 20 additional business days at no cost. If that is still not enough, we refund the sprint fee. This requires your timely cooperation; details are in the sprint offer.
Right of withdrawal: 4 weeks unconditional right of withdrawal with a full refund.
We run at most 3 Proof Sprints per month — we confirm the next available start date in the first call.
See pricing & Proof SprintFAQ
FAQ
Can NDA review AI replace a lawyer?
No. Treasure delivers a documented first review against rules your legal team has set and confirmed. Standard NDAs that meet every criterion can be cleared faster, while deviations go to a lawyer with a rationale and a suggested wording. Reviewers can add risks and give feedback, and the final decision stays with your team. Treasure does not provide legal advice.
Which NDA criteria can the AI check?
Any criterion your team can describe: mutual or one-way obligations, the definition of confidential information, confidentiality term, standard carve-outs, residuals, non-solicitation, contractual penalties, return or destruction of information, governing law and jurisdiction. You also set your party perspective, so the playbook knows whether you disclose, receive or both.
Why not just paste NDAs into ChatGPT or Claude?
A chat answer depends on the prompt, without a fixed NDA standard confirmed by legal and without a per-contract audit trail like Treasure's. Treasure applies the same confirmed NDA playbook every time, re-checks returned versions and stores results in the audit trail. If your team likes working in ChatGPT or Claude, the Treasure MCP server (AI tool integration, €50/month, included from Professional) lets them use Treasure from there.
Is it safe to send NDAs to Treasure by email?
Incoming emails are checked with DMARC and SPF, and only whitelisted senders per tenant are accepted. The platform runs on AWS in Frankfurt, AI processing happens in EU regions and customer data is not used for training. Information security is ISO 27001 certified.
What does NDA review with Treasure cost?
Plans start at €590 per month for Starter, with Professional at €1,790 and Business at €3,590, each net and billed annually; Enterprise is on request. The 10-day Proof Sprint costs €4,900 and is fully credited towards implementation if you book within 30 days. A 4-week right of withdrawal with full refund applies.
How long does it take to set up NDA review?
NDAs are usually the fastest contract type to start with because they are frequent and standardised. In the Proof Sprint, we derive the playbook from 10–20 historical NDAs and test it live within ten working days. If the acceptance criteria agreed in writing beforehand are not met, we add up to 20 working days at no cost and otherwise refund the sprint price; your cooperation is required.
Read next
Related solutions and features
Bring your next contract.
In 30 minutes you will know whether Treasure can apply your rules — on a real contract, with the founders.
Free · PDF guide
Implementing CLM with AI: A–Z
A practical workbook with ready-to-use prompts for requirements, vendor selection, rollout, adoption and ROI.
Double opt-in, unsubscribe anytime. No sales call, no spam.