3-Stage
Architecture.
AES-256 Encrypted.
API-first.

Algorithmic detection

Specialized language models

Isolated containers with customer separation

Architecture

The system processes documents in memory. Originals are not stored. After anonymization, only hash mappings exist for later restoration.

Document intake

API or web

Processing

Isolated containers

Hash storage

With customer isolation

Output

Anonymized

Detection methods

Pattern-based detection

Regular expressions and algorithms identify structured data:

  • +Phone numbers, email addresses
  • +IBAN, tax numbers
  • +Date formats, postal codes

Compact language model

A fine-tuned model detects:

  • +Names in various contexts
  • +Unstructured personal references
  • +Indirect identifiers

Context model

The larger model reviews the whole document:

  • +Validation of previous detections
  • +Contextual relationships
  • +Hidden references

Encryption

  • AES-256 for stored hash pairs
  • TLS 1.3 for transmissions

Data isolation

  • Customer A cannot access customer B's data
  • Separate encryption keys
  • API-key-based authentication

Retention periods

  • Configurable from 24 hours to unlimited
  • Automatic deletion after expiration
Encryption, data isolation and retention periods

Document processing

Large documents

Documents are split into sections (chunks). Each section is processed individually. The context model still considers the entire document.

Supported formats

  • −PDF (including scanned)
  • −Word, Excel
  • −Images (JPG, PNG)
  • −Plain text

Error handling

On processing errors, you receive a detailed error message. Partially processed documents are discarded.

Technical foundations

We use established components:

  • Modern language models (fine-tuned for data protection)
  • Standard OCR for text recognition
  • Proven encryption procedures

The uniqueness comes from:

  • Specific training of the models
  • Combination of the three detection stages
  • Integration with LLM instructions

API

Integration

The API uses JSON over HTTPS with Bearer token authentication.

Detailed documentation and endpoints are available in our API documentation.

Features

  • Synchronous and asynchronous processing
  • Webhook notifications
  • Fair use rate limiting
  • Comprehensive error handling

Start now.
14 days free trial.

All models. All features. No credit card.

Start for freeHow it works

We stand behind anymize. And we know – when an AI tool touches client, patient or employee data, a demo video isn't enough. That's why we give you 14 days of full access – all models, all features, no credit card. Enough time to be certain, before you trust us.

Your AI workplace awaits.