Snowflake PII Redaction

Redact PII where your data lives.

Call a single Snowflake function from the workflows your team already uses: dbt models, scheduled tasks, worksheets, stored procedures, notebooks, and pipelines. Redact text columns, support tickets, chat logs, PDFs, images, and DICOM without sending PII to an external API.

Start Snowflake Trial See If Agent Mask Fits

Marketplace is fastest if you are ready to try it. If you are not sure it fits, describe your workflow and we will tell you whether Agent Mask can help before you install.

Snowflake worksheet 
1SELECT agent_mask_en.app_public.mask(
2  [ticket_note, chat_log]
3) AS redacted
4FROM support_tickets;
Preview Same entity. Same replacement. Every column.
Input column Text
ticket_note Billing question from Maya Chen PERSON_1 about invoice INV-1042. Assigned to Alex Patel PERSON_2 . Contact: maya.chen@example.com EMAIL_ADDRESS_1 or 415-555-0198 PHONE_NUMBER_1
chat_log 10:42 Alex Patel PERSON_2 : Maya Chen PERSON_1 never got the invoice email.
10:43 Maya Chen PERSON_1 replied from maya.chen@example.com EMAIL_ADDRESS_1 and asked Alex Patel PERSON_2 to call 415-555-0198 PHONE_NUMBER_1

Built for the messy PII that Snowflake teams actually have.

Structured data masking protects known columns. Agent Mask is for the sensitive text and documents that do not fit neatly into column policies.

Free text

Redact names, emails, phone numbers, SSNs, addresses, IDs, and custom entities in tickets, notes, chats, call transcripts, and AI agent output.

Analytics-safe output

Choose the policy by entity type: preserve cardinality with deterministic hash or encryption, or fully replace values with constants, masks, labels, or faker values.

Consistent identities

Co-reference resolution links variants and fragments, so repeated people, organizations, and identifiers keep one replacement across columns.

Documents

Process PDFs, scanned PDFs, images, DICOM, DOCX, RTF, and ZIP archives inside your Snowflake account.

Audit trail included

Return each detected value, entity type, placeholder, and replacement alongside the de-identified text for review and downstream controls.

De-identification in one call.

Agent Mask gives data teams precise controls for different privacy goals without building a separate pipeline outside Snowflake.

Need
Agent Mask approach
Why it matters
Remove visible PII
Replace entities with `[PERSON_1]`, `[EMAIL_ADDRESS_1]`, `[REDACTED]`, or a configured value.
Keeps sensitive values out of downstream analytics, AI/ML training data, logs, and shared exports.
Control data utility
Configure each entity type to preserve cardinality with deterministic hash or encryption, or collapse values with constants, masks, and labels.
Lets privacy teams choose whether analytics keeps useful repeat-value signals or receives total replacement.
Handle aliases
Use entity collapse so "Sarah Chen," "Dr. Chen," and "Chen" share one placeholder.
Reduces inflated people counts and keeps long-form notes internally consistent.
Built-in option

Using Snowflake AI_REDACT?

AI_REDACT is useful when you need a built-in Cortex function for short, well-formed text. For larger production jobs, compare the documented limits with Agent Mask before building chunking logic around token caps, language coverage, and output constraints.

Compare AI_REDACT View all comparisons
Marketplace app

Need deeper redaction controls?

Agent Mask is built for teams that need document redaction, custom entities, multilingual detection, configurable de-identification output, and zero data egress from Snowflake.

Start Snowflake Trial See If Agent Mask Fits

Snowflake PII redaction FAQ

Questions data teams ask when they are choosing how to de-identify text and documents in Snowflake.

Should we use Snowflake AI_REDACT or Agent Mask?

Use AI_REDACT when a built-in Cortex function is enough for short, well-formed text. Use Agent Mask when the job needs document redaction, custom entities, multilingual detection, configurable redaction policies, consistent replacements, and metadata for review.

What Snowflake data can Agent Mask redact?

Agent Mask handles unstructured text columns such as tickets, notes, chats, call transcripts, intake fields, and AI agent output, plus supported files including PDFs, scanned PDFs, images, DICOM, DOCX, RTF, and ZIP archives.

Does Agent Mask send PII to an external API?

No. Agent Mask runs in the consumer Snowflake account using Snowpark Container Services. The masking service does not call external APIs, and there is no data ingress to Agent Mask infrastructure or egress from your Snowflake account.

Can we use different redaction policies for different entity types?

Yes. Teams can choose typed placeholders, constants, character masks, hashes, faker values, reversible encryption, or detect-only mode by entity type. That lets you preserve useful cardinality for some values while fully replacing others.

Can replacements stay consistent across related columns?

Yes. Agent Mask can keep repeated identities and aliases mapped to consistent replacements across related text, so downstream analytics does not lose every relationship in the data.

What evidence do we get back after redaction?

Responses include the de-identified output plus entity metadata such as detected entity type, detected value, placeholder, and replacement. That gives data and compliance teams something to inspect instead of only a final redacted string.

Get PII out of your working data.

Install Agent Mask from Snowflake Marketplace and run the first masking query inside your own account.

Start Snowflake Trial See If Agent Mask Fits