Shield applies persistent, field-level protection to sensitive data across your estate.
Shield protects the sensitive bit inside the file, whether it's the word, the field, the paragraph. Run it and watch what an attacker would actually get.
Most teams have classified their data. They have sensitivity labels, a posture score, a dashboard full of green. What they don't have is any assurance the data those labels point to is actually protected.
A file marked "Highly Confidential" in metadata, a ticket routed and queued and approved, and content still sitting in plaintext, exactly as exposed as it was the day it was tagged.
Shield reads the same label and acts on it. Classification becomes the trigger for cryptographic protection, applied at the field level, in the same platform, with no ticket in between.
Connect your sources, read the classification already in your environment, apply protection at scale, and keep control of every key, with no agents, and no re-tagging.
Point Shield at S3, SharePoint, OneDrive, Azure Blob, Box, and on-prem servers. Agentless for most cloud sources.
Whether it's identified or ingested, Discover is Shield's native analysis engine, scanning and classifying sensitive content directly, or reading existing labels from Purview, Varonis, or BigID. Either path feeds the same policy engine.
Encrypt, tokenize, redact, delete, or archive across the full estate, from terabyte to petabyte, at the granularity each field needs.
Revoke one person's access remotely without re-encrypting anything. Keys stay in your environment. Confidencial never sees plaintext.
It protects the sensitive element, not the whole file, so data stays usable, collaboration keeps working, and a breach produces nothing an attacker can read.
Protect at the word, field, paragraph, page, or full-document level. Original format is preserved, which means a Word doc still opens as Word.
Encrypt, tokenize, redact, delete, or archive, all chosen per data type, sensitivity, and who needs access. One policy engine decides.
Protects the full data estate. No per-GB caps, a 1TB job runs in roughly 13 hours.
Your KMS, your environment. Keys are identity-bound through your IDP, so you can revoke one person's access without
re-encrypting anything.
Turns Purview, Varonis, and BigID sensitivity labels into enforcement triggers. Add Discover for native classification.
Built for a seamless transition to NIST post-quantum standards. The data you encrypt today stays protected, no re-protecting the estate later.
Four situations where a label leaves you exposed, and where Shield changes what an attacker, an auditor, or an acquirer actually finds.
The download came from an authorized account. DLP didn't fire. Encryption-at-rest protected the bucket, not the file once it left it.
Every file is field-encrypted. They hold 3,000 files and can read none of the sensitive content. Access is revoked at the identity level with no device touch.
Eighteen months of labeling. Every dashboard green. But "Highly Confidential" is metadata, not a control on the content. Labels don't encrypt.
Shield reads the labels and encrypts every file they point to with no re-tagging. The classification investment finally drives enforcement.
The dataset has real value, such as outcomes, patterns, cost trajectories. It also contains PHI that can't touch an LLM without HIPAA exposure. The usual answer: don't use it.
PHI is encrypted at the word level. The model trains on everything else and hits ciphertext where PHI appears.
On-prem costs are mounting and AI initiatives are waiting. But every migration stalls on the same question: what happens to PHI, IP, and financials in a cloud bucket?
Protect the data before it migrates. Field-level encryption travels with it. The question becomes "is the data protected?", and it is.
File types protected, original format preserved
Data Solution of the Year in Legal, Data Breakthrough Awards
Tradewinds Marketplace Awardable
Purview classifies and labels your data — it doesn't protect it. A sensitivity label is metadata: it tells you a file is sensitive, but it doesn't encrypt the content. Shield reads your Purview labels and applies cryptographic protection to the files they identify, turning classification into actual enforcement.
Full-file encryption makes the entire document unusable unless decrypted in full. Selective encryption protects only the sensitive fields, words, or paragraphs inside the file — the rest stays readable, usable for collaboration, and processable by AI. Protection at the element level, not the file level.
The opposite — Shield is what makes AI workflows safe to run. Because protection is field-level, AI systems process the non-sensitive content freely and encounter ciphertext only where sensitive fields appear. You get the analytical value of your unstructured data without feeding PHI or IP into an LLM.
You do. Confidencial never sees plaintext. Keys are managed in your environment — bring your own KMS or use Shield's built-in key server. Access is identity-bound through your existing IDP, so revoking one person's access is an IDP group change, not a re-encryption job.
Yes. Shield ingests classification signals from any DSPM you already have — Purview, Varonis, BigID, Laminar — or you define protection policies directly by data type and regulatory category. Discover adds native classification if you want discovery and enforcement from one platform, but Shield operates independently.
Shield connects to your environment in days. No re-architecture. Just persistent protection for the sensitive data your classification tools have already found.