A model is still running

A model is still running or downloading. Leaving will stop the current operation. Stay on this page to let it finish.

Skip to content
LocalAI.toolsBeta

Join the waitlist

Get an email when the new apps and model runs launch.

No account needed.

Privacy Filter

Find and label spans in your text, processed locally in your browser.

Browser inferencePowered by WebAI · Private · Unlimited runs
About this model

Flag potential personal information in text across eight privacy categories. Returns labeled spans for a reviewed redaction workflow; missed or incorrect detections mean it cannot guarantee anonymization.

Created by
OpenAI
License
Apache-2.0
Model released
Unknown
Source repository created
Apr 17, 2026
Model weights · Q4
917.3 MB

Best for

  • Assist local or on-premises data-minimization and redaction workflows.
  • Find candidate PII spans for review before text is shared or indexed.
  • Evaluate privacy-filtering behavior against an organization's in-domain policy and data.

Keep in mind

  • This is a detection aid, not an anonymization, compliance, or safety guarantee; false negatives may expose sensitive data and false positives may remove necessary context.
  • It detects only its fixed taxonomy: account number, address, email, person, phone, URL, date, and secret.
  • Performance can drop on non-English text, non-Latin scripts, uncommon names, regional formats, domain-specific identifiers, long documents, and heavy layout or punctuation.
  • Secret detection may miss novel credentials or flag benign high-entropy text.
  • The browser artifacts are very large: approximately 809 MB to 5.63 GB of weights before shared tokenizer files. The verified Q4 audit profile did not retain its weights in Chrome Cache API, so affected sessions redownload them.
  • High-sensitivity medical, legal, financial, employment, education, and government uses require in-domain evaluation and human review.

Processed on your deviceInference runs in this browser. Input/output sharing requires consent for each run. Basic run logs include device information and approximate connection location. Model and runtime files download from external hosts. Saved downloads may be cleared by your browser.

Sources & technical details
Original model
openai/privacy-filter
Release verification
Unknown
Release evidence
Exact original checkpoint public release date has not been verified. Repository creation, training, paper, artifact conversion and update dates are not release evidence.
Source updated
Apr 22, 2026
Browser model files
openai/privacy-filter
Files updated
Apr 22, 2026
Artifact revision
7ffa9a043d54d1be65afb281eddf0ffbe629385b

Input

TEXT → ENTITIES

Sample result

OpenAI Privacy Filter — PII Span Detection
Advanced settings 1
How adjacent token predictions are combined into privacy spans. none keeps individual tokens. simple groups compatible tokens; first uses the first subtoken label, average combines subtoken scores, and max uses the strongest subtoken. Available choices can change entity boundaries and labels.

Output

Discover the labels in your text

Enter text or try the sample to highlight entities and inspect their scores.

Benchmark evidence

Saved worker evidence is separate from benchmarks run on your device. It does not guarantee speed or quality on every browser.

OpenAI Privacy Filter — PII Span Detection

10 configurations tested
Passed2
Failed8
TestedSep 15, 2026
RuntimeTransformers.js 4.2.0

Quality criteria

Finite spans must detect the fictional name Harry Potter as private_person and harry.potter@hogwarts.edu as private_email at score >= 0.9, with no more than one unexpected span and valid offsets when present.

PrecisionDeviceResult
Q4GPU · WebGPUAcceleratedPassedFast
Q4F16GPU · WebGPUAcceleratedPassedModerate
Q4CPU · WASMFailed—
Q4F16CPU · WASMFailed—
Q8CPU · WASMFailed—
Q8GPU · WebGPUFailed—
FP16CPU · WASMFailed—
FP16GPU · WebGPUFailed—
FP32CPU · WASMFailed—
FP32GPU · WebGPUFailed—
Test setup & sample details
Quality gate
Finite spans must detect the fictional name Harry Potter as private_person and harry.potter@hogwarts.edu as private_email at score >= 0.9, with no more than one unexpected span and valid offsets when present.
Tested
2026-09-15T12:41:45.169Z
Environment
Isolated headless Chrome on macOS; Transformers.js 4.2.0
Runtime
Transformers.js 4.2.0
Artifact revision
7ffa9a043d54d1be65afb281eddf0ffbe629385b
Fixture
{
  "file": "privacy-filter.json",
  "url": "https://f005.backblazeb2.com/file/hubters-webai-public/assets/benchmarks/privacy-filter.json",
  "size": 899,
  "sha1": "39b9188d3224c759479738aa156538e01a283414",
  "source": "Official OpenAI Privacy Filter model-card Transformers.js example.",
  "sourceUrl": "https://huggingface.co/openai/privacy-filter/blob/7ffa9a043d54d1be65afb281eddf0ffbe629385b/README.md#transformersjs",
  "license": "Apache-2.0",
  "licenseUrl": "https://github.com/openai/privacy-filter/blob/f7f00ca7fb869683eb732c010299d901457f19c3/LICENSE",
  "preprocessing": "Pinned tokenizer followed by constrained model-specific sequence decoding and simple span aggregation."
}
Rerun when
artifact revision change; runtime version change; fixture or quality-gate change; browser/backend behavior change
Per-case quality & failure details
[
  {
    "precision": "q4",
    "device": "wasm",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": {
      "loadMs": null,
      "inferenceMs": null,
      "totalMs": 31244.40000000596,
      "grade": "failed"
    },
    "failure": "Error: Can't create a session. ERROR_CODE: 9, ERROR_MESSAGE: Could not find an implementation for GatherBlockQuantized(1) node with name '/model/embed_tokens/Gather_Quant'"
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "finite": true,
      "validOffsets": true,
      "expectedMatches": [
        true,
        true
      ],
      "unexpectedEntities": 0,
      "grade": "privacy-span-pass"
    },
    "runtime": {
      "loadMs": 31144.5,
      "inferenceMs": 166.2000000178814,
      "totalMs": 31406.5,
      "grade": "good",
      "charactersPerSecond": 397.1119133146756
    }
  },
  {
    "precision": "q4f16",
    "device": "wasm",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": {
      "loadMs": null,
      "inferenceMs": null,
      "totalMs": 29940.5,
      "grade": "failed"
    },
    "failure": "Error: Can't create a session. ERROR_CODE: 9, ERROR_MESSAGE: Failed to find kernel for com.microsoft.GatherBlockQuantized(1) (node:'/model/embed_tokens/Gather_Quant' ep:'CPUExecutionProvider'). Kernel not found"
  },
  {
    "precision": "q4f16",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "finite": true,
      "validOffsets": true,
      "expectedMatches": [
        true,
        true
      ],
      "unexpectedEntities": 0,
      "grade": "privacy-span-pass"
    },
    "runtime": {
      "loadMs": 30789.5,
      "inferenceMs": 638.5,
      "totalMs": 31589.899999976158,
      "grade": "slow",
      "charactersPerSecond": 103.3672670321065
    }
  },
  {
    "precision": "q8",
    "device": "wasm",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": {
      "loadMs": null,
      "inferenceMs": null,
      "totalMs": 52944.90000000596,
      "grade": "failed"
    },
    "failure": "Error: Can't create a session. ERROR_CODE: 9, ERROR_MESSAGE: Could not find an implementation for GatherBlockQuantized(1) node with name '/model/embed_tokens/Gather_Quant'"
  },
  {
    "precision": "q8",
    "device": "webgpu",
    "status": "fail",
    "quality": {
      "finite": true,
      "validOffsets": true,
      "expectedMatches": [
        false,
        false
      ],
      "unexpectedEntities": 9,
      "grade": "failed"
    },
    "runtime": {
      "loadMs": 54158.70000001788,
      "inferenceMs": 635.5,
      "totalMs": 54955.90000000596,
      "grade": "slow",
      "charactersPerSecond": 103.8552321007081
    }
  },
  {
    "precision": "fp16",
    "device": "wasm",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": {
      "loadMs": null,
      "inferenceMs": null,
      "totalMs": 88576.5,
      "grade": "failed"
    },
    "failure": "Error: Can't create a session. ERROR_CODE: 6, ERROR_MESSAGE: std::bad_alloc"
  },
  {
    "precision": "fp16",
    "device": "webgpu",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": {
      "loadMs": null,
      "inferenceMs": null,
      "totalMs": 85955.40000000596,
      "grade": "failed"
    },
    "failure": "Error: Can't create a session. ERROR_CODE: 6, ERROR_MESSAGE: std::bad_alloc"
  },
  {
    "precision": "fp32",
    "device": "wasm",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": {
      "loadMs": null,
      "inferenceMs": null,
      "totalMs": 144554.5,
      "grade": "failed"
    },
    "failure": "Error: Can't create a session. ERROR_CODE: 6, ERROR_MESSAGE: std::bad_alloc"
  },
  {
    "precision": "fp32",
    "device": "webgpu",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": {
      "loadMs": null,
      "inferenceMs": null,
      "totalMs": 143015,
      "grade": "failed"
    },
    "failure": "Error: Can't create a session. ERROR_CODE: 6, ERROR_MESSAGE: std::bad_alloc"
  }
]

Something not quite right?

Tell us what happened with Privacy Filter. Loading, results, or the page itself—we’re here for all of it.

Attachments Optional

Up to 3 files · 50 MB each

Browser, OS, CPU/GPU, memory, screen, storage, and network details are included automatically. Model inputs, results, and browsing history are not collected.