openai/privacy-filterBidirectional token-classification model for PII detection and masking in text, with CLI for redaction, evaluation, and finetuning on-premises.
OpenAI Privacy Filter is a bidirectional token-classification model for personally identifiable information (PII) detection and masking in text. It is intended for high-throughput data sanitization workflows where teams need a model that they can run on-premises that is fast, context-aware, and tunable.
OpenAI Privacy Filter is pretrained autoregressively to arrive at a checkpoint with similar architecture to gpt-oss, albeit of a smaller size. We then converted that checkpoint into a bidirectional token classifier over a privacy label taxonomy, and post-trained with a supervised classification loss. (For architecture details about gpt-oss, please see the gpt-oss model card.) Instead of generating text token-by-token, this model labels an input sequence in a single forward pass, then decodes coherent spans with a constrained Viterbi procedure. For each input token, the model predicts a probability distribution over the label taxonomy which consists of 8 output categories described below.
Highlights:
This repository contains the local code, CLI, and example assets used to run, evaluate, and finetune Privacy Filter checkpoints. It is meant for teams that want to inspect the implementation directly and operate the model in their own environment.
Repository resources: License and Security Policy.
pip install -e .
After this, you will have a python script opf that can be run directly or via python -m opf. The script can be used in 3 separate ways, as described below.
By default, opf looks for a model at the directory pointed to by the OPF_CHECKPOINT variable, or ~/.opf/privacy_filter. If a model is not found in the ~/.opf/privacy_filter location, it will be downloaded.
opf "Alice was born on 1990-01-02."
The code supports running both on GPU (by default) and CPU. To run on CPU, use --device cpu flag:
opf --device cpu "Alice was born on 1990-01-02."
To override the default checkpoint, pass --checkpoint:
opf --checkpoint /path/to/checkpoint_dir "Alice was born on 1990-01-02."
The redaction mode supports redacting an entire file at once
opf -f /path/to/file
The redaction can also be performed via pipes, to support complex one-liners:
cat /path/to/file | grep -e 'some_pattern' | opf
If no input is provided, opf will start in interactive mode. In this mode, for each input example, the CLI prints structured JSON output, using ANSI color-coded previews if the terminal supports them. These options can be controlled by flags.
Consult opf redact --help for more flags and information about the redaction mode.
opf eval examples/data/sample_eval_five_examples.jsonl
The sample eval fixtures under examples/data/sample_eval_five_examples*.jsonl are synthetic example data only and do not describe real people or real sensitive records. See examples/data/README.md.
Consult opf eval --help for more flags and information about the evaluation mode.
opf train /path/to/train.jsonl --output-dir /path/to/finetuned_checkpoint
Consult opf train --help for more flags and information about the finetuning mode.
opf/__main__.py: unified CLI entrypoint for redact, eval, and train modes.opf/_api.py: Python-facing API over the runtime and decoding stack.opf/_cli/: command-line argument parsing and terminal rendering helpers.opf/_core/: runtime loading, span conversion, and shared decoding logic.opf/_eval/: dataset loading, preprocessing, metrics, and evaluation runners.opf/_train/: local finetuning argument parsing and training runners.opf/_model/: transformer implementation, checkpoint config, and weight loading.examples/data/: sample eval files plus reproducible finetuning demo datasets.examples/scripts/finetuning/: runnable finetuning demo harnesses.FINETUNING.md: focused finetuning workflow and demo-script guide.OUTPUT_SCHEMAS.md: JSON response and export payload formats.EVAL_AND_OUTPUT_MODES.md: description of the output modes for redaction and evaluation.Privacy Filter is a bidirectional token classification model with span decoding. It is trained in phases, beginning with autoregressive pretraining. The pretrained language model is then modified and post-trained as a bidirectional banded attention token classifier with band size 128 (effective attention window: 257 tokens including self). This means:
Architecturally, the implementation in this repo is a pre-norm transformer encoder-style stack with:
d_model = 640.Relative to iterative autoregressive approaches, this design allows all tokens to be labeled in one pass, which improves throughput. Relative to classical masked-language-model pretraining approaches, this is a post-training conversion of an autoregressive model rather than a native masked-LM setup.
Privacy Filter can detect 8 privacy span categories:
account_numberprivate_addressprivate_emailprivate_personprivate_phoneprivate_urlprivate_datesecretTo perform token-classification, each non-background span category is expanded into boundary-tagged token classes: B-<label>, I-<label>, E-<label>, S-<label>, plus the background class, O. So the total number of token-level output classes is 33: 1 background class + 8 span labels * 4 boundary tags = 33 classes. This means the output head emits 33 logits for each token. For a sequence of length T, the output has shape [T, 33]; for a batch of size B, it has shape [B, T, 33].
The token-label vocabulary consists of the background label O plus BIOES-tagged variants of each privacy category: account_number, private_address, private_email, private_person, private_phone, private_url, private_date, and secret. In other words, for each category, the model predicts B-, I-, E-, and S- forms corresponding to begin, inside, end, and single-token spans. At inference time, these per-token logits are decoded into coherent BIOES span labels using constrained sequence decoding.