Privilege ends where a public LLM begins.
Drop a draft contract, pleading, or correspondence into the editor. We flag client names, opposing counsel, addresses, SIN, file numbers, and settlement amounts. Accept what you want, redact in one pass, and either share the static PDF or send the tokenized text into Claude or ChatGPT for a second read.
- Detection covers client and counsel names, addresses, SIN, postal codes, phone, email, and account numbers.
- Reversible tokenization recovers the original from the redacted file with a passphrase only you hold.
- Bilingual at the engine. French pleadings and English contracts run the same model.
- PIPEDA and Quebec Law 25 documentation provided on enterprise request.
PHI never reaches the summariser.
Clinical notes, discharge summaries, and referral letters are the files most likely to end up in a chat window at 7 a.m. We strip names, dates of birth, health card numbers, addresses, and phone numbers before any AI pass, and keep the clinical context the model actually needs.
- Strips names, DOB, health card numbers (RAMQ, OHIP, etc), phone, email, and addresses.
- Reversible only by the clinician who holds the passphrase. The redacted PDF can leave the device.
- HIPAA aligned controls, signed BAA, on prem LLM bundle available in the enterprise tier.
Need HIPAA controls and a signed BAA? Talk to enterprise
ATIP, briefing notes, and the Privacy Act.
Federal and provincial teams handle documents where a misrouted line draws scrutiny. We run in the browser under the same rules in English and French, hold nothing of the file binary, and ship with Canadian residency on every byte we do hold.
- Bilingual detection on the same engine. Output parity guaranteed.
- Document binary never reaches our server. Detection, redaction, and key handling run in the browser.
- Canadian residency on the roadmap for every byte we hold (auth, billing metadata).
- Privacy Act and Quebec Law 25 documentation provided on request.
Need audit log export or Privacy Act documentation? Talk to enterprise
Score the qualifications. Not the candidate.
We strip identifying detail from a stack of CVs before they reach an AI ranker. Names, contact details, postal codes, and graduation dates come out. Education, experience, and skills stay in. Reversible tokenization brings the candidate back when it is time to make an offer.
- Batch processing for an entire CV pile in one session.
- Custom rules for the fields your ATS exports.
- Reversible tokenization. The candidate returns when the decision needs a real name.
- Bias mitigation as a byproduct of the redaction, not an extra workflow.
Reuse the analysis. Not the client identity.
Take a deliverable from one engagement and make it shareable for the next pitch. We remove client names, project codes, internal contact details, and identifying numbers while keeping the structure, charts, and argument intact.
- Strips client names, addresses, contact details, and internal identifiers.
- Multi PDF batches for sanitising an entire engagement folder at once.
- Reversible tokenization for the consultant. Static PDF for the external audience.
SIN, account numbers, and T-slips before the AI pass.
Tax returns, account statements, and KYC files carry the densest concentration of identifiers in any document on the desk. We strip SIN, account and routing numbers, T4 and T1 detail, and personal contact information before a single line reaches an AI tool.
- Detects SIN, credit card numbers, account numbers, postal codes, and contact detail.
- Reusable custom rules for the fields your firm processes most often.
- Reversible tokenization keeps client review possible without exposing identifiers.
- PIPEDA and Quebec Law 25 documentation provided on enterprise request.