piiguard[border]

The piiguard detector for border, an embeddable library that inspects the text going into and coming out of an LLM and returns a structured decision plus an audit-grade evidence record.

flowxai/piiguard on the hub. It is one detector of 28, and it is not a general purpose piiguard classifier: it was trained for this library's policy, is read at the operating point below, and reports through the evidence record rather than returning a bare score.

This card is generated from the evaluation and export artifacts of the training run, so every number on it is reproducible from this repository rather than asserted.

What it is

• Base model: FacebookAI/xlm-roberta-base
• Head: token classification, BIO tags
• Labels: O, B-PERSON, I-PERSON, B-EMAIL, I-EMAIL, B-PHONE, I-PHONE, B-NATIONAL_ID, I-NATIONAL_ID, B-IBAN, I-IBAN, B-CARD, I-CARD, B-DATE, I-DATE, B-LOCATION, I-LOCATION
• Artifact: onnx/model.fp16.onnx, 555 MB, opset 17
• Trained at: 96 tokens

Operating point

This head is read with argmax and has no threshold.

How to use it

Through the library, which is what this model is for. It loads the artifact below, applies the operating point above, and returns a decision with an evidence record rather than a bare score.

pip install flowx-border

# policy.yaml
policy_id: default
version: 1

detectors:
  piiguard:
    enabled: true
    on_fail: flag

from flowx_border import load_policy, scan_input, scan_output

policy = load_policy("policy.yaml")

decision = scan_input(user_text, policy)
decision = scan_output(model_answer, policy)

print(decision.verdict)      # allow | flag | redact | block
print([f.label for f in decision.findings if f.detector_id == "piiguard"])
print(decision.evidence.record_id)

This detector reads the input and output side, so scan_input and scan_output is where it fires. It is T2, so it runs on the standard path and can be disabled per policy.

The weights are fetched once and cached, and a scan needs no network after that. Nothing here calls out to a hosted model, and the evidence record carries hashes rather than your text.

Without the library

The artifact is plain ONNX, so it will load in onnxruntime directly. Two things you then own yourself, and they are the reason the library exists: the operating point above is not in the graph, and neither is the chunking. Inputs longer than the trained window have to be split and recombined, or the scores past it are extrapolation.

import onnxruntime as ort
from huggingface_hub import hf_hub_download
from tokenizers import Tokenizer

repo = "flowxai/piiguard"
session = ort.InferenceSession(hf_hub_download(repo, "onnx/model.int8.onnx"))
tokenizer = Tokenizer.from_file(hf_hub_download(repo, "tokenizer.json"))

Per language

Per language rather than an aggregate, because an aggregate across 26 languages hides the tail and the tail is the point.

Language	Support	P	R	F1	Note
az Azerbaijani	164	1.000	1.000	1.000
bg Bulgarian	180	1.000	1.000	1.000
da Danish	124	1.000	1.000	1.000
de German	212	1.000	1.000	1.000
el Greek	144	1.000	1.000	1.000
en English	180	1.000	1.000	1.000
es Spanish	180	1.000	1.000	1.000
et Estonian	188	1.000	1.000	1.000
fi Finnish	164	1.000	1.000	1.000
hr Croatian	152	1.000	1.000	1.000
hu Hungarian	224	1.000	1.000	1.000
lt Lithuanian	148	1.000	1.000	1.000
lv Latvian	156	1.000	1.000	1.000
mt Maltese	164	1.000	1.000	1.000	not in base model pretraining
nl Dutch	160	1.000	1.000	1.000
pl Polish	180	1.000	1.000	1.000
pt Portuguese	172	1.000	1.000	1.000
ro Romanian	192	1.000	1.000	1.000
sk Slovak	188	1.000	1.000	1.000
sl Slovenian	152	1.000	1.000	1.000
sv Swedish	132	1.000	1.000	1.000
tr Turkish	172	0.994	0.994	0.994
it Italian	172	0.988	0.988	0.988
cs Czech	164	0.976	1.000	0.988
ga Irish	160	0.976	1.000	0.988
fr French	148	0.987	0.987	0.987

Weakest languages

Published rather than dropped. A coverage table with the bad rows removed is not a coverage table.

• fr French: F1 0.987
• ga Irish: F1 0.988
• cs Czech: F1 0.988

Quantisation

The published artifact is fp16, halving every weight rather than quantising a subset. Used where INT8 moved decisions on this base model and fp16 did not.

For this artifact specifically: ? of 300 decisions differ from the fp32 checkpoint, mean logit drift not recorded, read as character spans. A quantised model that answers differently is a different detector, so this is measured rather than assumed.

Limitations

• Synthetic training data. Generated natively per language, never translated from English, so the sentence structure is the target language's own. It is still synthetic, and a production distribution will differ.
• Maltese is absent from XLM-RoBERTa's pretraining set. That is a fact about the base model, and it is not an explanation for a weak score. This card said "no amount of data fixes that" until 2026-08-14, which this project's own measurement disproves: the nsfw detector scored 0.000 in Maltese, was blamed on the base model, and went to 1.000 with perfect precision and recall when its corpus went from 2 positives per language to 10. Nothing about the model changed. So where a language scores badly here, read the support column first.
• This is not a compliance product. It produces evidence about controls that were applied. It does not make anyone compliant with anything, and the obligations under the EU AI Act sit with the provider or deployer of a system, not with a model or a library.

Licence

Apache-2.0, declared in the metadata above as well as here, so that a tool reading the repository can attest it rather than a human having to read prose.