Zero data retention
Prompts and documents are never stored off-device
Local AES-256 encryption
Full-disk encryption at rest, keys held by the firm
SOC 2 Type II · HIPAA
Built for attorney-client privilege obligations
Offline processing
Runs fully air-gapped — no egress, no API calls
Vault AI — the on-premise legal appliance
A sealed AI appliance that sits in your office and travels with you — so privileged work never touches a cloud provider.
100% Private
Client files never leave the firm — the unit holds the model and the matter data on local encrypted disk.
Zero Cloud Leakage
No internet connection required. Nothing is uploaded, so there is no third-party retention to review.
Plug-and-Play
Arrives configured. Plug it in, add your team, and start work the same afternoon — no IT project.
Free Hardware Pilot • Priority Batch Allocation • Dedicated On-Site Support
Less time on document review and case prep
Of routine matters resolved on-premise
Bytes of client content retained by any model
Our risk committee approved it in one sitting. The appliance never opens an outbound connection, so there was nothing left to argue about.
Deposition prep that used to eat a full afternoon now comes back in about twenty minutes — with citations we can actually check.
Every run is logged, privileged material stays inside our network, and the cost is a fixed line item. That combination did not exist before.
Reviewed against SOC 2 Type II, ISO 27001, HIPAA and ABA Model Rule 1.6 confidentiality obligations.
Every workflow runs entirely on the appliance — no document ever leaves the building.
Redline agreements against your own playbook in a single pass.
3 hrs → 15 min
First-pass redline per agreement
Find the authority and the responsive passages without a weekend of review.
4 hrs → 20 min
Deposition and hearing prep
Clear a data room or a discovery batch before the client calls back.
2 weeks → 1 day
Full diligence pass
Existing platforms like Harvey and CoCounsel put your client data in their cloud, meter every query, and bill per attorney every month. Vault AI is the opposite — an on-premise appliance you own, with no tokens to rent, no seats to pay for, and no data that ever leaves the building.
Hosted in a vendor cloud. Client data leaves your building on every query.
100% on-premise. Zero bytes of client data leave the appliance — ever.
$1,000–$1,500+ per user, per month. The invoice climbs with every attorney.
$999/month for the entire firm + one-time $2,999 appliance. Flat, forever.
Every new lawyer adds a recurring seat fee — growth gets expensive.
Unlimited attorney seats. Adding lawyers costs nothing.
You rent frontier tokens (GPT, Claude) — metered and metered again.
You own the local models. Every query resolves on the box — $0 token cost.
Dead without an internet connection to the vendor.
Works fully offline. No internet required for any legal task.
Chat history sitting in someone else's database.
Immutable, matter-level audit trail — hashed, timestamped, on your disk.
Non-cancellable contracts. They can raise the price at renewal.
You own the box. The advantage doesn't expire with a contract.
One box you own. One flat fee. Zero client data in motion. That's the entire pitch — and it's everything the cloud alternatives can't give you.
Three commitments we put in writing, and an audit trail that proves them after the fact.
The appliance sits inside your firm's perimeter. Every matter is resolved entirely on the box — no outbound connection, no vendor account, no third-party processor in the path.
Nothing your attorneys submit is used for training, fine-tuning or evaluation. The local model weights are fixed on your hardware and learn nothing from your matters.
Privileged content never leaves the building. Matters are scoped and access-tiered on the appliance, and working memory is cleared the moment a run completes.
Each matter is walled off with its own access tier and encrypted store, so a conflict wall in the firm is a conflict wall in the appliance — no shared index, no cross-matter leakage.
A predictable firm-wide subscription — a one-time appliance plus a flat monthly fee that keeps the box licensed, updated, and supported, with no per-query metering and no per-seat penalties.
For boutique, mid-tier, and enterprise legal departments requiring complete data containment.
The monthly fee keeps the appliance evolving, compliant, and protected — so your firm always runs on current models and a healthy box.
Regular pushes of faster, smarter, state-of-the-art open-source LLM weights.
Automatic monthly updates to local statutory databases and court rules.
100% hardware replacement coverage if the box fails, plus guaranteed hardware upgrades every 2 years.
Outbound-only security updates to keep your local environment air-tight.
Cloud legal AI (CoCounsel, Harvey) runs $225–$1,200 per user, per month. Vault AI is $999/month for the entire firm — plus a one-time $2,999 appliance — with unlimited seats on every box.
A 10-lawyer firm on $225/seat cloud AI spends $27,000/year. Vault AI costs ~$15,000 in Year 1 and only ~$12,000/year after — under $100 per lawyer per month.
The appliance turns a variable API bill into a single fixed line item — the same amount every month, however heavily your teams use it.
Work on the appliance is unmetered — one flat firm-wide fee, no per-query metering, no per-seat penalties, and no external API bill.
Hold a unit from our next hardware batch. Tell us about your firm and we'll schedule your pilot and on-site install.
Fully refundable deposit • No hardware lock-in
