Browser DLP for AI Tools
Stop sensitive data from leaving through the AI chat tab.
Prism X enforces data loss prevention policy in the browser, before prompts and uploads reach third-party AI services. Signed policy, real-time enforcement, audit-grade events.
- Pre-send
- enforcement before egress
- Ed25519
- signed policy + lineage
- 4 tools
- ChatGPT, Claude, Gemini, Copilot
- MDM-ready
- Intune, Google, Jamf
Capabilities
Everything DLP needs for the AI chat surface
Pre-send enforcement, signed policy, MDM rollout, and audit-grade events, built for the channel network DLP was never designed to inspect.
Pre-send DLP
Content is evaluated and enforced before the browser dispatches the request. Network DLP never sees these prompts; Prism X does.
DLP rule library
Pattern-based and contextual detection for PII, PHI, credentials, and confidential markers, with validators that reduce false positives.
Signed policy engine
Cryptographically signed policy with version lineage and real-time distribution. Tamper-resistant enforcement, not 'we have a policy.'
Managed deployment
Push policy keys, API endpoints, and trust material via Microsoft Intune, Google Admin, Jamf, or your existing MDM.
Multi-tool coverage
ChatGPT, Claude, Gemini, Microsoft Copilot, plus new AI tools that follow standard chat input patterns, often without explicit per-site configuration.
Admin dashboard
Live activity, fleet health, event drill-down, and policy management. Role-scoped so the right people see the right data.
Audit events
Structured events for every enforcement decision, policy change, device registration, and admin action. Append-only, exportable, framework-mappable.
Why it matters
Network DLP doesn't see prompts. CASB doesn't inspect chat input. Prism X covers the channel that your existing stack was never designed for.
Pre-send enforcement
Evaluated and decided before the request leaves the browser.
A browser-level content layer attaches to AI chat input surfaces and file upload controls. When the user clicks Send, the policy engine evaluates content in priority order, before the browser dispatches the request.
- Allow, warn, block, or redact per rule
- DOM-bound integration, not network interception
- Resilient to vendor UI updates
- Covers ChatGPT, Claude, Gemini, Copilot, plus standard chat surfaces
Detection catalog
Rules that match what regulated data actually looks like in the wild.
Pattern-based and contextual detection with validators that reduce false positives. Six built-in categories covering credentials, financial, government, healthcare, and legal markers, plus column-header-aware scanning for CSV and XLSX uploads.
- Six categories ship out of the box, all with validators
- Column-header-aware spreadsheet scanning
- Five rule types: domain, file-type, keyword, regex, contextual
- Priority-ordered evaluation with explainable outcomes
Signed policy engine
Policy you can trust, version, and prove.
Cryptographically signed policy with version lineage and real-time distribution. The extension verifies the signature on every poll. If verification fails, it falls back to the last known good policy rather than going unprotected.
- Ed25519-signed policy verified on every poll
- 30-second propagation, push-style for urgent rollouts
- Version lineage: who, when, what, version number
- Last-known-good fallback on signature failure
Managed deployment
Deploy to your fleet the way your IT team already works.
Push policy keys, API endpoints, and trust material to managed browsers via Microsoft Intune, Google Workspace Admin, or Jamf. Force-installed, force-configured, with fleet visibility and last-known-good fallback.
- Microsoft Intune, Google Admin, Jamf, Firefox enterprise policies
- Three-step rollout: package, push, register
- Fleet view: enrollment, policy currency, coverage percentage
- Per-org Ed25519 public key, cross-tenant injection impossible
Audit events
Every decision, every timestamp, every policy version, exportable.
Append-only events covering enforcement, policy changes, device registrations, fetches, and admin actions. Privacy controls let your legal team set user-identification granularity. Sensitive content redacted at source, never reaches the backend.
- Five event types covering enforcement, policy, device, admin
- Configurable user-identification granularity
- Append-only with retention to match your framework
- Redacted snippets only, raw sensitive content never stored
Why Prism X
Built for the channel your CASB and SWG can't inspect.
Pre-send, not post-send
Content evaluated before the browser dispatches the request. Network DLP only ever sees what already left.
Signed, versioned, tamper-resistant
Ed25519-signed policy verified on every poll. Last-known-good fallback if verification fails.
MDM-native
Force-installed via Intune, Google Admin, or Jamf. Employees can't disable, uninstall, or misconfigure.
Audit-grade events
Append-only events with privacy controls, exportable for SOC review and compliance evidence packages.
Covered AI tools
Where your workforce actually uses AI.
DOM-bound integration covers the chat surfaces in real production use: ChatGPT, Claude, Gemini, and Microsoft Copilot. New AI tools that follow standard input patterns are often covered without per-site config.
See all supported toolsPrompt scanning, file-upload extraction, share-link DLP
Prompt scanning, project-knowledge upload control
Prompt scanning, file-upload extraction
Prompt scanning across Edge and Chrome
Aligned to the frameworks your auditors reference
Frequently asked
Common questions
How does Prism X differ from a CASB or network DLP?▾
CASB and network DLP inspect file transfers, email, and API traffic. They don't see what an employee types into a chat box. Prism X attaches at the browser, scanning prompts and file uploads at the moment of submit, so you cover the channel those tools were never designed for.
Which AI tools are supported?▾
ChatGPT (chat.openai.com and chatgpt.com), Anthropic Claude (claude.ai), Google Gemini (gemini.google.com), and Microsoft Copilot (copilot.microsoft.com) ship in the box. New AI tools that follow standard chat input patterns are often covered without per-site configuration.
How is Prism X deployed?▾
Force-installed and force-configured via Microsoft Intune, Google Workspace Admin, Jamf, or Firefox enterprise policies. The extension reads its config from managed storage, so employees can't disable, uninstall, or misconfigure. Three-step rollout: package, push, register.
What happens if a policy update fails to verify?▾
The extension falls back to the last verified policy rather than going unprotected or accepting tampered configuration. Every policy is Ed25519-signed by your tenant; the extension verifies the signature on every poll.
What does the audit log capture?▾
Five event types: enforcement decisions, policy changes, device registrations, policy fetches, and administrative actions. Each event carries timestamp, AI destination, rule, pattern category, disposition, and policy version. Sensitive content is redacted at source, never stored on the backend.
Cover the AI data risk your stack doesn't.
Pre-send DLP for the four AI tools your workforce actually uses, with signed policy, MDM rollout, and audit-grade events. Deploy in a day.