Services · Guardrail models

Guardrail models built on your data, small enough for every call.

We build small, dense guardrail models for the risks you care about, from your own data, with the same recipe as our own family. Each is fast enough for the request path and portable enough to run anywhere, and it's yours.

Our guardrail model family

  1. Personal data, v0.1Finds and masks 77 kinds of personal data in English.
    Open weights
  2. Personal data, v0.2Adds XML, SQL, log and key-value traffic to prose.
    Open weights
  3. Personal data, multilingualThe same job across languages.
    Built in-house
  4. Named entitiesPeople, organisations and places in free text.
    Built in-house
  5. ToxicityAbuse and harm in what people and agents send.
    Built in-house
  6. Prompt injectionAttempts to override instructions, including disguised ones.
    Built in-house

Six models, two with open weights on Hugging Face.

The recipe

Small, dense, fast and portable.

A guardrail sits in front of every AI call, so it has to cost almost nothing. That decides how we build: a small model made for one job, not a large general one asked to behave.

Small

Built from scratch for one job, with its own tokenizer and architecture, not a general model cut down. Our open personal-data model is 43 MB.

Dense

One pass over the text gives several decisions at once, each from its own small head on a shared encoder, so a check costs one read, not several.

Fast

Made for the request path, where every call waits for the check: 1.4 ms median per check on an ordinary 8-core CPU.

Portable

Each model is one ONNX file. It runs on a CPU, at the edge or fully air-gapped, and in VidaiGuard beside the Vidai gateway.

The evidence

Measured in public.

Our open personal-data model is ranked against every model submitted to the PIIMB benchmark, large and small.
3rd of 36on the PIIMB English leaderboard, 0.014 behind a 1.4-billion-parameter leader*
43 MBthe whole model, one ONNX file*
1.4 msmedian per check on an ordinary 8-core CPU, no GPU*
77kinds of personal data found and masked*

*PIIMB English leaderboard, 17 October 2026: Stirling-Minimo PII v0.1 ranked 3 of 36 at 0.904 average masking F2 across four English sets, 0.014 behind the leader, a 1.4-billion-parameter model. Size, speed and coverage from the Stirling-Minimo PII model card; on a GPU it serves 7,600 checks a second on one previous-generation GPU, using about 3 GB of its 24 GB.

Build yours

A model for your risk, trained on your data.

Generic safety models are trained on someone else's idea of harm. Yours should know your customers' data, your products' names, your languages and the formats your systems actually send.
  1. 1

    Define the risk

    What must be caught, in which languages and formats, and what a miss costs you. Your policy, not a generic safety list.

  2. 2

    Gather the data

    Your own examples, labelled with you, plus generated cases for what's rare. We can work inside your environment, so the data never has to leave it.

  3. 3

    Train from the recipe

    A small model built for that one job, from the recipe behind our own family.

  4. 4

    Prove it

    Measured on your held-out data and on public benchmarks where one exists, with a regression suite that runs on every retrain.

  5. 5

    Ship it

    One ONNX file, running in VidaiGuard beside the gateway or anywhere ONNX runs.

What you own

Everything, with no dependency on us.

The model

The weights and the ONNX file, under your control. No call back to us to run it.

The pipeline

The training and data preparation code, in your repository, so you can retrain as your data changes.

The evidence

The evaluation suite and its results, to show an auditor or a risk committee what the model catches and what it misses.

Where it helps most: regulated data in healthcare, finance and legal work; languages and formats a general model handles badly; and anywhere a check must run without a GPU or a network.

40+years of combined engineering, built at the scale you operate at

Production systems shipped at

Amazon · GE · Tata

and other operators, not slide decks.

How we start together

Bring us the risk you need to catch.

A direct conversation with the engineers who build the models. Tell us what must not get through, and in what.