How to Safely Use AI with Sensitive Data

The Core Principle

Most data protection tools work by securing data after it leaves your control. Redactli works by ensuring sensitive data never leaves your browser in the first place.

When you upload a CSV to Redactli:

1. Your file stays local. The raw CSV is processed entirely in your browser.
2. Only column values are sent for anonymization. Not your file. Not the structure. Just the specific values that need protection.
3. You receive anonymized data back. Sensitive values are replaced with human-readable tokens like 'Name_a3b7c9d2', 'Email_k9m2p4x7', making it easy to reference specific entries in AI conversations.
4. You use the safe version with any AI tool. Paste it into ChatGPT, Claude, or any other service without risk.

What Gets Anonymized (and How)

Redactli applies different protection methods based on data type:

Names → Human-Readable Tokens

Example: "Sarah Chen" becomes "Name_a3b7c9d2"

Encrypted values formatted as readable tokens, making it easy to reference specific people in AI conversations.

Email Addresses → Tokenized Format

Example: "sarah.chen@company.com" becomes "Email_k9m2p4x7"

Encrypted as readable tokens while keeping identity fully protected.

Phone Numbers → Tokenized Values

Example: "+61 412 345 678" becomes "Phone_m7n3q8r2"

Encrypted as human-readable tokens for easy reference.

Addresses → Tokenized Format

Example: "42 Collins Street, Melbourne" becomes "Address_x5y9z3w1"

Encrypted as readable tokens while protecting actual locations.

Financial Data → Transformed Values

Account numbers, credit card numbers, and financial identifiers are replaced with valid-format alternatives that protect actual account details.

Custom Fields → Flexible Handling

You choose which columns to anonymize. Leave non-sensitive data unchanged while protecting PII columns.

Why Anonymization, Not Encryption?

Encryption protects data from being read, but AI tools need to read your data to be useful. Encrypted data is meaningless to ChatGPT.

Anonymization takes a different approach: it transforms data so it can be read and analysed, but the real identities are protected.

Deterministic Anonymization

A key feature of Redactli's approach is deterministic anonymization. This means the same input always produces the same output.

Why does this matter?

1. Consistency across analyses

If "John Smith" appears 47 times in your dataset, he becomes "Robert Johnson" 47 times. Your analysis remains valid.

2. Referential integrity

Foreign key relationships survive anonymization. Customer IDs in your orders table still match customer IDs in your customers table.

Step-by-Step Workflow

How Redactli protects your data in practice:

Step-by-Step: Protecting Your Data

Follow these steps to safely use AI with sensitive data:

Step 1: Upload Your CSV File

Click "Upload CSV" or drag and drop your file. Redactli uses Papa Parse to process your CSV entirely in your browser—no data is sent to our servers at this stage.

Step 2: Select Columns to Anonymize

Review your column headers. Redactli automatically suggests columns that appear to contain PII (names, emails, phone numbers, addresses). Select or deselect as needed.

Step 3: Anonymize the Data

Click "Anonymize" to transform selected columns. Only the values from these columns are sent to our secure API for encryption using AES-SIV. The encrypted values are returned as human-readable tokens like 'Name_a3b7c9d2' and 'Email_k9m2p4x7', then merged back into your CSV client-side.

Step 4: Download and Use with AI

Download your anonymized CSV. You can now safely upload it to ChatGPT, Claude, Google Bard, or any AI tool. The AI will see realistic data while real identities remain protected.