Anaris Labs
Use everything. Share nothing.
Anaris Labs is the data protection company for Canadian work. We build tools that keep personal information off the wires you don't control. Files stay on your device, on your network, or with you.
No more private data on the web where it isn't needed.
Share without leaking.
We find the sensitive parts of your files automatically. Hand off a contract, screenshot, or PDF without reading every line first.
Reversible only by you.
Everyone sees black bars. Only you, and the people you choose, can lift them and recover what's underneath.
AI keeps the context.
Swap names, numbers, and IDs for stable placeholders. Your AI tools keep the context they need and never see the originals.
One protection model. Three places it runs.
Same detection model across every surface. What changes is how you reach it.
A.Redactextension
Research PreviewAutomatically remove personal data before uploading files anywhere online.
- Custom rules, reversible tokenization, and AI prompts like
remove all mention of Sarah's birthday.
Free while in Research Preview.
Learn moreA.Redactweb
Research PreviewClick and drag editor with multi-PDF batches.
Custom rules, reversible tokenization, and configurable AI prompts.
- Get started without an account.
Free while in Research Preview.
A.Veil
Coming SoonNetwork-level redaction for enterprise systems.
- Intercepts at the edge. No change in user behaviour.
- HIPAA, SSO, BAA, on prem LLM, audit log export.
Enterprise preview, talk to us.
Learn moreReal workflows. Real documents. No leaks.
For the kind of work where one misrouted document is more expensive than any tool. Works on the files you use everyday.
Lawyers
PIPEDA · Quebec Law 25 · by architecture
Client privilege ends the moment a draft contract reaches a public LLM.
The draft never leaves your machine, so privilege survives the second read. We strip client names, opposing counsel, addresses, file numbers, and settlement amounts before anything reaches Claude or ChatGPT.
Read the full caseHealthcare
HIPAA · PIPEDA · PHIPA aligned
Patient data leaks because no one wants to read 30 pages at 7 a.m.
We strip names, dates of birth, health card numbers, and addresses before they reach your AI summarizer, and keep enough clinical context for the model to actually be useful at the bedside.
Learn moreRecruiters
PIPEDA · Quebec Law 25 · bias mitigation
A stack of CVs, an AI ranker, and a candidate privacy obligation you signed three weeks ago.
Score the qualifications without seeing the candidate. Reversible tokenization brings the real name back the moment you need it.
Learn morePublic sector
Privacy Act · Quebec Law 25 · Canadian residency
ATIP requests, briefing notes, internal memos. The wrong line in the wrong place is a federal headache.
English and French run the same detection model, so output parity is built in rather than negotiated. Canadian residency on every byte we hold, with documentation on request.
Learn more
Posture, not promises.
Six controls every privacy product wishes it could ship. They hold because the architecture, not the marketing, does the work.
Built for the regulated.
HIPAA aligned controls, signed BAA, on prem LLM bundle, SSO, audit log export. Available in the enterprise tier.
Bilingual at the engine.
Detectors work the same on English and French, with more languages coming soon.
Canadian residency.
Built under PIPEDA and Quebec Law 25. The little we hold, like auth and billing metadata, is on a roadmap to Canadian processors. Documentation provided on enterprise request.
Reversible tokenization.
Tokenized redactions recover with a passphrase only the holder knows. The key derivation and cipher choices are public; the passphrase is yours.
Local LLM models.
AI augmentation is provider-agnostic. Point it at Cohere, Anthropic, or a model running on your own hardware. Air-gapped environments stay air-gapped.
Browser first protection.
The document binary never reaches our server. Detection, redaction, tokenization, and encryption all run on the client. AI augmentation, when used, only sees extracted text and can run on a local or on-prem model.
Stop sharing data you don't need to.
Free while in Research Preview. No account needed to start.