Browser AI Security Standard
Purpose: Set security requirements for users accessing browser-based AI platforms such as public or enterprise generative AI services.
Audience: Security engineering, DLP teams, endpoint teams, SOC, AI users, compliance.
| Control Area | Control ID | Weight | Required Evidence |
|---|---|---|---|
| Discovery and Visibility | AI-CTRL-001 | 15% | AI service inventory, user activity reports, device attribution |
| Prompt Inspection | AI-CTRL-002 | 20% | Sensitive prompt detection logs and policy outcomes |
| File Upload Inspection | AI-CTRL-003 | 15% | Upload inspection test results and block/warn evidence |
| Policy Enforcement | AI-CTRL-010 | 15% | Allow, alert, warn, mask, redact or block policy evidence |
| Auditability | AI-CTRL-008 | 15% | User, device, AI platform, prompt and file upload logs |
| Platform Coverage | N/A | 10% | Browser and OS support matrix |
| Operational Impact | N/A | 10% | Deployment notes, performance observations, administrative effort |
Minimum Requirements
- Discover AI platforms accessed from managed endpoints.
- Identify users, devices and browser context.
- Inspect prompts where technically feasible.
- Inspect uploaded files including documents, spreadsheets, PDFs, images and code files where technically feasible.
- Apply classification-aware monitor/warn/redact/block controls.
- Generate investigation-ready audit records.
- Forward AI security events to SOC/SIEM.
PoC Tests
- Upload confidential document to public AI platform.
- Paste architecture details into AI prompt.
- Upload source code to AI chatbot.
- Attempt upload of classified information.
- Access newly emerging or previously unknown AI website.
Audit Evidence
- AI service inventory export.
- Screenshots of policy enforcement decisions.
- Logs showing user, endpoint, browser, AI site, action and timestamp.
- SIEM event ID or SOC alert reference.