Senior-grade contract review in 11 minutes, with attorney-client privilege intact.
Every NDA, MSA, SaaS agreement and M&A draft your team uploads is processed inside a hardware-sealed Intel TDX enclave on European infrastructure. Privileged client data never reaches a third-party model provider, and your work product never trains anyone else. Trained on the same review protocols used by Magic Circle and Am Law 100 firms.
Built for: Law firms 5-50 lawyers, in-house legal teams, procurement directors
Manual contract review takes 4-6 hours per agreement. Outside counsel charges $500-1,500/hr. ChatGPT and most legal AI tools process privileged client data on shared infrastructure, breaching attorney-client privilege the moment you paste a draft.
Cut contract review from 4 hours to 11 minutes per agreement, under hardware-rooted confidentiality.
What the Contract Analyst does on every document, sealed inside an Intel TDX hardware enclave.
Black / Red / Amber / Green ratings on every clause, with quoted source text and section numbers. Reserved BLACK ratings flag genuinely deal-breaking provisions (unlimited liability, perpetual non-competes, asymmetric IP assignment).
Compares the document against IACCM, ABA Model Rules and procurement-team checklists. Surfaces protections you would expect to see but do not, with draft language you can paste back.
For every flagged clause, returns proposed replacement language formatted as a tracked-changes redline, plus a fallback position if the primary ask is rejected.
Drop in an MSA and an SOW, or a CIM and a financial annex; the agent flags inconsistencies between documents instead of reviewing each in isolation.
Knows the difference between Delaware common law, English law, French civil law and Singapore-governed APAC SaaS. Calls out terms that are unenforceable in the chosen jurisdiction.
Extracts every deadline, notice requirement, delivery date and renewal trigger into a structured table you can hand to ops or paste into a CLM tool.
Four steps from upload to export. Your document is decrypted only inside the CPU-encrypted enclave.
Drag in PDFs, DOCX or scanned contracts. Files are transmitted over TLS and decrypted only inside the Intel TDX enclave on the GPU host. The disk is never touched.
The Qwen3-32B-TEE model runs inside an Intel TDX hardware enclave. Memory is AES-encrypted by the CPU; even the cloud operator cannot read the prompt or the document.
Executive summary, key commercial terms table, risk findings with section citations, missing protections, and an obligation tracker. Every claim is traceable to a section number.
Download the redline package as Markdown or DOCX, paste into Word with track-changes enabled, or pipe the JSON into your CLM via API (Pro tier and above).
Every output cites the specific article, section or control. The Intel TDX hardware boundary handles the data-protection side; the agent handles the analytical side.
Real prompts from legal teams.
Review this NDA — I'm the receiving party. Flag anything non-standard.
Is the limitation of liability in Section 8 market-standard for B2B SaaS?
Extract all obligations with deadlines from this MSA.
Compare the indemnification in this SOW vs the MSA — are there conflicts?
Prompt: Rate the overall risk of signing this vendor MSA as-is.
Output: RED — Asymmetric liability (Sec. 11.2: vendor capped at fees paid; you uncapped on IP indemnity). Unilateral IP assignment of derivatives (Sec. 14.4) — non-standard for an MSA, this is an IP assignment clause hiding in a services agreement. Recommend redline before signature.
Prompt: What clauses are missing that a procurement team would expect?
Output: Missing: (1) Mutual termination for material breach with cure period; (2) Data Processing Addendum despite Sec. 6 referencing personal data; (3) Service credits / SLA remedies — Sec. 9 promises 99.9% uptime with no financial consequence on miss; (4) Source code escrow for critical-path SaaS.
Every tier runs inside Intel TDX hardware enclaves. Plans stay in sync with /pricing.
For solo practitioners. 1,000 requests/month, 1 seat, 25 MB uploads. Always-warm Intel TDX inference.
3 seats, 500 requests/month, 100 MB uploads, audit log. Designed for small firms billing $300-800/hr.
10 seats, 5,000 requests/month, 500 MB uploads, full API access for CLM integration, audit log retained 12 months.
Unlimited seats and requests, fine-tuning on your firm’s clause library, SSO/SAML, DPA, custom DPIA, dedicated TDX capacity.
Honest comparison. Hardware-rooted confidentiality is what most alternatives are missing.
| Alternative | Pros | Cons vs VoltageGPU |
|---|---|---|
| Spellbook |
|
|
| Harvey AI |
|
|
| CoCounsel (Thomson Reuters) |
|
|
No. Inference runs on an open-weight model (Qwen3-32B) deployed inside an Intel TDX hardware enclave on European infrastructure. The prompt, the document and the response never leave the encrypted memory of the GPU host. No data is sent to OpenAI, Anthropic, Google or any other third-party model provider, and nothing is used for training.
Qwen3-32B running inside an Intel TDX enclave on the Starter tier; the larger Qwen3.5-397B-A17B model on Pro and Personal Agent tiers; and DeepSeek-R1 reasoning on Enterprise. All deployments are hardware-attested with Intel DCAP.
No. Privilege is broken when privileged content is disclosed to a third party. Because the Intel TDX enclave is a hardware boundary — not a software promise — the cloud operator cannot read the data even during computation. This is the same architecture cited by the EU Data Act and California state bar guidance on confidential AI use.
Yes, on the Enterprise tier. We support fine-tuning on your historical redlines and clause precedents, plus custom prompt templates per matter type. The fine-tune itself is performed inside a confidential VM and the resulting weights are kept private to your tenant.
The underlying confidential compute platform is operated by VOLTAGE EI (France, SIREN 943 808 824). Coverage: GDPR Art. 28 DPA (signed on request), SOC 2 Type II in audit, ISO 27001 roadmap. Intel TDX provides hardware-rooted attestation independent of any software certification.
Yes, starting on the Pro tier ($1,199/mo). The API is OpenAI-compatible — you change the base URL, keep your existing SDK, and the request runs inside the enclave. This is how customers integrate the analyst into their CLM (Ironclad, ContractWorks, Agiloft, Lexion).
A single contract review at $500-1,500/hr from a senior associate runs $2,000-6,000 per agreement. The Starter tier handles roughly 50 contracts/month for $349 — a 100x reduction in marginal cost. Outside counsel is still appropriate for negotiation strategy and litigation; the analyst handles the first-pass review.
The output is framed as “analysis for review by qualified counsel,” not as legal advice. It is designed to surface what a senior associate would surface on a first pass, so partner time is spent on judgment calls rather than line-by-line reading.
Intel TDX attestation, EU jurisdiction, French operator (VOLTAGE EI). Cancel anytime.