TEGORAI

Governance · Security · Trust

The house that keeps your data under cover.

Tegorai lets regulated professions and organisations use AI without exposing their sensitive information: data is protected on your side, before it reaches the AI model.

Discover ArtagentiaJoin the MVP programme

Closed beta from 27 November 2026 · launch in January 2027

The house

Tegere: to cover, to protect.

Our name comes from the Latin tegere, to cover. Like a roof sheltering a house, Tegorai places a solid structure over your data and your digital operations.

Our symbol is a builder’s flat arch: three stones and a central keystone that holds them together. That is how we work — simple parts, precisely assembled, locked by one principle: you stay in control.

Tegorai is a company based in the canton of Vaud, founded by partners from cybersecurity, enterprise architecture, private banking and financial management.

The need

AI adoption is moving faster than control over data.

Banning AI breeds workarounds. Adopting it without safeguards exposes professional secrecy, personal data and know-how.

  • The user wants to keep their tools and save time, without changing the way they work.
  • The data may contain names, health records, contract clauses, identifiers or internal information.
  • The organisation must be able to explain what is allowed, blocked, transformed and logged.
  • The proof must still hold years later: documents, contracts and operations attested without exposing their content, and protected against future quantum computers.

Since 2 August 2026, the transparency obligations of the EU AI Act apply. Together with the Swiss FADP, NIS2, DORA and FINMA requirements, the direction is the same: you must be able to prove, not just declare.

Three functions

One structure, three roles.

  1. 01

    Protect

    Keep data and exchanges under cover: what leaves your organisation is controlled, and what must stay confidential does.

    • Artagentia — data protection Beta 2026
  2. 02

    Govern

    Supervise AI at work: budgets, explicit rules and human approval of irreversible actions. Assess AI vendors before entrusting them with data.

    • Artagentia — AI agent supervision Beta 2026
    • GRC assessment of AI vendors In preparation
  3. 03

    Attest

    Prove what was done: reliable, verifiable records, ready for the auditor, the regulator or the court.

    • CrystalMint project — trust registry Late 2027 – 2028

Artagentia · a Tegorai product

Protect data before it leaves your hands.

Artagentia is the protection layer that sits between your AI tools and the provider you have chosen. You keep working as before; what must not leave, does not leave.

Control stays with you. The choice of model stays open.

  1. 01

    Detect

    As you write in your usual tool, sensitive elements are identified — on your side.

  2. 02

    Decide

    Your policy applies: allow, replace, block, or process the most sensitive data on your side.

  3. 03

    Replace

    Sensitive values become tokens. Only the protected text goes to the AI model you chose.

  4. 04

    Restore

    The answer comes back and the original values are restored on your side, for you alone.

21 types of sensitive data
identity, contact, finance, secrets
22 languages
including French, German and Italian
99.5% detection score
F1 on our internal bench
98.3% health data
F1, dedicated detection (GDPR Art. 9)
25 ms median added delay
for 512 characters · 52 ms for 1,024

Internal measurements on synthetic test corpora (1,478 samples in 22 languages; 557 for health data), summer 2026. The delay depends on the equipment and the length of the text. Results vary by language and data type; the beta will test them against real-world use.

“Where is your text pseudonymised — on your side, or at yet another provider?”
Many offerings promise sovereignty while routing your data through their own cloud. Artagentia answers “on your side”: the original value is never handed to an intermediary.

Commitments

Write as you would a contract.

We prefer verifiable commitments to slogans. Here is what we stand by.

On your side by default
Detection, decision and restoration happen on your side. Your values are never replaced at an intermediary.
Your keys stay yours
You keep your models and your own access keys. The mapping table stays with you, under your control.
No telemetry
The software sends no usage data to Tegorai.
People decide
AI proposes, a person approves. Irreversible actions always require explicit approval.
Proof before promise
We separate what is delivered and measured from what is planned, and publish the conditions of our measurements.
Compliance by design
Designed for the Swiss FADP, the GDPR, the EU AI Act, DORA and NIS2: a tool that helps you prove, not automatic compliance.

Three editions, one software

Community

Personal use

Free for personal, non-commercial use: core protection.

January 2027

Pro

Professionals and SMEs, from one seat

Full protection including health data, processing of the most sensitive data on your side, support within two working days.

January 2027

Enterprise

Organisations, from 100 seats

Central governance, compliance attestation, audit log for your security tools.

Planned for 2027

Works with your usual AI tools and your own access keys

Timeline

  1. 27 Nov 2026

    Closed beta

    Testers under non-disclosure agreement.

  2. Jan 2027

    Artagentia Pro 1.0

    Online purchase, installation and first protected text in a few minutes.

  3. 2027

    Enterprise

    With design partners: central governance and integration with security tools.

Working objectives, not contractual commitments. A date is not a proof: each stage is only passed when tests and real use confirm it.

CrystalMint project · late 2027 – 2028

A trust infrastructure for regulated sectors.

CrystalMint is Tegorai’s distributed registry project. It gives every sensitive document, contract or operation a unique, verifiable fingerprint, with a history designed so that it cannot be rewritten — without exposing its content. Smart contracts execute the rules agreed between the parties (conditions, deadlines, approvals) in a traceable way.

Data is protected by the post-quantum algorithms standardised by NIST in 2024, against attacks that collect encrypted data today in order to decrypt it tomorrow. Threshold access lets a judge or a regulator consult a piece of information for a limited time, with the agreement of several parties and without ever exposing the master key.

Our post-quantum approach

Future quantum computers threaten the algorithms that protect banking, medical and legal archives today — archives that are sometimes kept for decades. We already apply the standards published by NIST in August 2024, following four principles.

  1. 01

    One algorithm per role

    ML-KEM (FIPS 203) for confidentiality, ML-DSA (FIPS 204) for signatures, SHAKE-256 (FIPS 202) for fingerprints — never one in place of another.

  2. 02

    Protection matched to lifetime

    NIST security level 3 by default; level 5 for data retained for more than thirty years.

  3. 03

    Hybrid during the transition

    Server-to-server connections combine a classical and a post-quantum algorithm: they stay secure as long as either one holds.

  4. 04

    Keys under hardware custody

    Target architecture: the most sensitive keys protected by hardware security modules (HSMs) hosted in Switzerland.

Our algorithms follow the standards published by NIST; this is not a FIPS 140-3 certification of the product, which we do not claim. Artagentia relies on the same standards (ML-KEM-768, ML-DSA-65).

Project under development: availability targeted for late 2027 – 2028. Presented for information.

Contact

Let’s talk about your project.

A question, a demonstration, a partnership? Write to us — we reply personally.

Registered office
Pully, Vaud, Switzerland
Business address
Trust Village
Route des Flumeaux 46
1008 Prilly, Vaud, Switzerland
From 3 November 2026