VaultLMRequest preview
Data shielded · sources attached · safe sharing · audit trail

Use AI with confidential documents. Without sharing raw data with an LLM.

VaultLM shields sensitive information before AI processing, keeps answers linked to sources and makes anonymised sharing safe. Use AI on documents you would normally never paste into a chatbot.

Request previewView demo
Data shielded
Sources attached
Safe sharing
Audit trail
Product demo

See what happens before AI sees your document.

Upload a confidential file, ask a question, see which data was shielded and verify the answer with sources.

VaultLM
Trust details on

Confidential dossier 2026

Only relevant, shielded passages go to the selected AI model. Original files and original text stay in the EU.

EU endpointno raw personal datacitations on
3 sources
selected sources · safe route
Proposition

The safe layer between your confidential documents and AI.

Upload a document, let VaultLM mask sensitive details and then ask questions with an LLM. You get useful answers with citations, without giving the model unnecessary access to raw confidential data.

01

Analyse confidential documents with AI

Upload contracts, student data, notes, reports or proposals and ask questions as naturally as in a normal AI chat.

02

Sensitive data shielded first

Names, emails, numbers, amounts, IBANs and other sensitive values are replaced before AI works with the content.

03

Answers with evidence

See exactly which passages support an answer, what information was shared and what was shielded.

How it works

From confidential document to safe AI answer.

Every question follows the same safe route: upload, shield, ask and review what was used.

01

Upload your document

Place your file in a secure environment and choose which sources may be used for your AI question.

02

VaultLM detects sensitive information

Names, contact details, numbers and other sensitive details are masked before AI works with them.

03

Ask your AI question

Only relevant, shielded context goes to the selected AI model. Not the entire raw document.

04

Review the answer

See sources, citations and which information was or was not shared with the model.

Safe sharing

Share a version people can actually receive.

Turn a confidential document into a clean, anonymised version for a client, investor, colleague or external partner — without exposing the original data.

Save
Masked copy

Store a version with sensitive details removed, while the original stays protected.

Share
Safe share link

Share a masked file with expiry, role, watermark, download policy and a reviewable log.

Export
Reviewable package

Export PDF, DOCX, TXT or ZIP with citations and clear metadata about what was shared.

Model integration

The AI gives the answer. VaultLM controls what the AI may see.

VaultLM sits between your documents and the AI model. Original files, original text, permissions and re-identification data stay in the EU; only shielded, relevant passages are used.

Default
Managed API via VaultLM

Use GPT, Claude, Gemini or Mistral without managing your own keys. VaultLM forwards only anonymised context.

Strict
Own API, EU endpoint or private model

Use your own API contracts, EU-resident endpoints, private cloud or self-hosted models.

Working context

Continue the work without pretending AI remembers everything.

Working context is a compact summary per user and secure workspace. It helps follow-up questions, but never replaces the original documents.

  • Included: goals, choices, definitions, preferences and conclusions.
  • Not included: full source documents or uncontrolled hidden memory.
  • Manageable: visible, resettable and policy-gated per user or workspace.
Control evidence

Not just useful. Also reviewable.

VaultLM helps you use AI without unclear data flows, loose prompts and answers nobody can trace back to their sources.

  • Traceability: every answer connects back to source passages.
  • Transparency: user, role, model, context and citations are logged.
  • Data minimisation: only relevant masked passages are used, not full raw files.
Important nuance

Reviewable is not the same as legally certified.

VaultLM provides architecture and evidence for controlled AI use, but does not replace formal legal, privacy or AI Act advice.

Product principleA user should never retrieve more through AI chat than they are allowed to open in the original documents.
Security

The assurance a standalone chatbot cannot give.

Sensitive-data shielding and access control are built into the route to the model, not left to a setting someone can forget.

A

Raw files stay protected

Original documents and original text remain encrypted in the EU. Only relevant masked passages are forwarded to the AI model.

B

Permissions stay in place

AI retrieval respects the same rights as document access.

C

Clear operational controls

Model choice, retention, log export and EU-only requirements can be explicitly documented.

Who uses VaultLM

For anyone who wants to use AI when the information is sensitive.

A teacher with student data. A journalist with source material. A founder with an investor deck. A lawyer with client files. A consultant with customer data. VaultLM makes AI usable when the information should not simply become public.

Teacher

Analyse student records, reports or assessment notes without sharing names and other sensitive details unnecessarily.

Journalist

Organise source material, transcripts and dossiers while names, locations or source details stay shielded.

Founder

Review pitch decks, contracts or investor documents without sending company-sensitive details straight to an LLM.

Lawyer

Compare client files, contracts and correspondence with citations and control over what is shared.

Consultant

Work with customer documents and reports without spreading raw client data outside a safe workflow.

Pricing

Start protected. Scale when it works.

Protected preview available without download or sharing rights. Annual plans include a 30-day money-back guarantee.

Personal

€9.50
per month annually · €19 monthly
  • 10 GB encrypted storage
  • Questions within fair use
  • Full anonymisation workflow
  • 30-day review log
  • Protected preview without download or sharing
Request preview

Team

€39
per user / month annually · €49 monthly
  • 100 GB shared storage
  • Role-based access
  • Saved anonymised files
  • Safe sharing packages
  • 1-year review log
Start pilot

Organisation

Custom
SSO · policy · EU route
  • Unlimited workspaces
  • Own API, EU route or private model
  • Advanced safe sharing policy
  • Custom retention and log export
  • Security and DPA review
Contact us


Warning: Undefined variable $pricingNoteText in /var/www/html/index.php on line 746

Q&A

Frequently asked questions.

Short answers about privacy, what data goes to AI, reviewability and safe sharing.

Does my original document go to GPT or Claude?

No. VaultLM does not send raw documents to the model. Only minimal, relevant and masked context goes to the selected AI route.

Does everything stay in Europe?

Original files, original text, token maps and re-identification data stay in the EU and never leave the EU. The selected AI route determines where shielded, minimised context is processed; EU-only routes are available.

Can masked files be shared?

Yes. Safe sharing packages can include expiry, role, download policy, watermark and a reviewable log.

Why not just anonymise manually and paste?

Manual anonymisation is error-prone. VaultLM makes shielding, citations and logging part of the same workflow.

Contact

Want to use AI on sensitive information? Request access.

Use this form for a demo, access, security questions, privacy agreements, safe sharing or pricing.

hi@vaultlm.aiEU-firstSecurity review ready

Privacy & security

How VaultLM handles your documents and data

VaultLM is built for people who want to use AI with confidential information, without that information simply ending up with an external model. That principle is built into the architecture, not only into the privacy policy.

What happens to your files

Uploaded files are stored encrypted in an EU-based secure environment. VaultLM never sends a complete raw document or original text to an AI model or outside the EU. Instead, VaultLM detects sensitive information — names, email addresses, phone numbers, customer numbers, amounts, IBANs and other identifying details — and replaces it with stable temporary tokens before a passage is sent to the model. The link between a token and the original value remains encrypted inside VaultLM and does not leave the secure environment.

AI-model answers are connected back to the passages they are based on. You can see which context was used, what information was shielded and through which route the question was processed.

What data VaultLM processes

VaultLM processes the files you upload, the questions you ask, the settings you choose and the account data required for access and billing. VaultLM does not use your files or the content of your questions to train models.

AI model providers

VaultLM works with external AI model providers such as OpenAI, Anthropic and Google. These providers receive only the shielded, minimised context required to answer your question — not your original files, original text, token maps or re-identification data. Original data stays in the EU and never leaves the EU. EU-only and self-hosted model routes are available for organisations that also want model processing to stay in the EU.

Retention

Files and log data are retained while your account is active, or as agreed in a data processing agreement. The Personal plan includes a 30-day log; Team includes 1 year. Custom retention periods are available for organisations.

Access and permissions

Access to files and secure workspaces is determined by the permissions set when uploading. VaultLM employees do not have access to the contents of your files unless you explicitly grant permission for technical support.

Your rights

You can delete your files, close your account or request an overview of the data VaultLM holds about you at any time. For privacy questions and requests, email hi@vaultlm.ai.

Applicable law

VaultLM operates from the European Union and uses the GDPR as the basis for data processing. The exact processing basis, data flows and subprocessors are described in the data processing agreement available on request.

DPA / security review

For organisations reviewing VaultLM before rollout

If you want to use VaultLM for team usage, customer-related workflows or processing personal data on behalf of an organisation, a formal review is often required. VaultLM actively supports that process.

Data Processing Agreement (DPA)

VaultLM acts as a processor under the GDPR when you upload and process personal data through the platform. A data processing agreement defines which data is processed, for what purpose, for how long, which subprocessors are involved and which safeguards are in place to protect data-subject rights.

A standard DPA is available on request. For organisations with specific requirements — different retention periods, additional instructions on model routes, restrictions on subprocessors or requirements around cross-border transfers — the agreement can be adapted.

Security review

VaultLM can support procurement, privacy and security teams with the information needed for a vendor assessment. Depending on your organisation’s review process, VaultLM can provide documentation about data flows and which information leaves the secure environment at which point in the processing.

VaultLM can also provide an overview of subprocessors, including AI model providers, and the agreements made with them. Information is available on encryption at rest and in transit, access controls, log management, incident response, available model routes, EU endpoints and self-hosted options. VaultLM can also help complete security questionnaires or vendor assessment forms.

Pilot setup

For organisations that want to test VaultLM in a bounded environment before full rollout, VaultLM can provide a guided pilot setup combining a specific model route, a limited user group and a tailored log configuration.

How to request a review

Email hi@vaultlm.ai with a short description of your organisation, intended use and preferred deployment route: managed API, EU endpoint or self-hosted model. VaultLM will respond within two business days to align the next steps.

Terms

VaultLM terms of use

Version May 2026

What VaultLM is

VaultLM is a platform for secure AI-assisted document analysis, shielding sensitive data and controlled sharing of anonymised document versions. VaultLM provides a technical layer that shields confidential information before an AI model can access it.

Your responsibilities

You are responsible for the files you upload and the questions you ask. This means you only upload files for which you have the required rights or permission, you review AI answers against the sources they are based on before using them externally, and you do not use VaultLM to process special categories of personal data — such as medical or criminal data — without a data processing agreement in place.

Limits of AI output

AI answers are based on the context sent to the model and may contain errors, omit information or misinterpret passages. VaultLM is not a substitute for legal, medical, financial or other professional advice. Always check the cited sources before making decisions based on an AI answer.

Liability

VaultLM provides the technical infrastructure and is responsible for the correct functioning of the shielding and routing mechanism. VaultLM is not liable for the content of AI answers, errors in detecting sensitive information, decisions made based on AI output, or damage resulting from incorrect or unlawful use of the platform.

Intellectual property

The files you upload remain yours. VaultLM uses the contents of your files only to provide the service to you. VaultLM does not use your files or questions to train AI models.

Accounts and access

You are responsible for keeping your login details secure. If you suspect misuse, contact hi@vaultlm.ai immediately. VaultLM reserves the right to block access in case of confirmed misuse or use that violates these terms.

Changes

VaultLM may update these terms. For material changes, you will be notified by email or at login. Continued use after the effective date of a change counts as acceptance.

Commercial agreements

Subscription terms, data processing agreements and service arrangements for organisations are recorded in a separate agreement. Use the contact form or email hi@vaultlm.ai to request an agreement.

Applicable law

These terms are governed by Dutch law. Disputes are submitted to the competent court in the Netherlands.