Skip to content

DLP Controls

Data Loss Prevention (DLP) controls help prevent accidental or intentional exposure of sensitive data through Fp Switchboard tools.

DLP in Fp Switchboard works by:

  1. Scanning tool inputs for sensitive data patterns
  2. Scanning tool outputs before returning to the AI
  3. Taking action based on your configured rules

PII Detection

Social Security Numbers, credit cards, phone numbers

Secrets Detection

API keys, passwords, tokens

Custom Patterns

Regex patterns for your industry or company

Keyword Blocking

Block specific terms or phrases

  1. Go to Admin → DLP Settings

    Navigate to switchboard.fpdigital.ai/admin/dlp

  2. Enable DLP scanning

  3. Select detection categories

  4. Configure actions

  5. Save and activate

CategoryExamplesDefault Action
SSN123-45-6789Redact
Credit Card4111-1111-1111-1111Block
Phone Number(555) 123-4567Log
Email Addressuser@domain.comLog
API Keysk_live_xxx, AKIA…Redact
Passwordpassword=, pwd:Block

Add your own regex patterns:

name: "Internal Project Codes"
pattern: "PROJ-[A-Z]{3}-\d{4}"
action: notify
description: "Detected internal project code"

Block or flag specific terms:

name: "Confidential Terms"
keywords:
- "confidential"
- "internal only"
- "do not share"
action: notify
ActionBehavior
BlockPrevents the tool call entirely
RedactReplaces sensitive data with [REDACTED]
NotifyAllows but alerts admins
LogAllows with enhanced audit logging

Different actions for inputs vs outputs:

detector: credit_card
input_action: block # Don't let AI send credit card numbers
output_action: redact # Redact if they appear in results

Some users may need to work with sensitive data:

exemptions:
users:
- finance@company.com
- hr@company.com
detectors:
- ssn
- credit_card

Some tools legitimately need access:

exemptions:
tools:
- quickbooks_create_invoice
- hubspot_create_contact
detectors:
- phone_number
- email

View DLP activity in Admin → DLP Reports:

  • Detection Summary — What types of data are being detected
  • Trend Analysis — Are detections increasing over time?
  • User Breakdown — Which users trigger the most detections
  • False Positives — Review and tune your detectors

If a detector triggers too often on non-sensitive data:

  1. Review the detection log
  2. Add exclusion patterns
  3. Adjust confidence thresholds
detector: phone_number
min_confidence: 0.8
exclusions:
- "555-" # Fictional phone numbers

Some patterns have confidence scores. Set minimum thresholds:

detector: credit_card
min_confidence: 0.95 # High confidence required
action: block

DLP logs integrate with your compliance requirements:

  • HIPAA — PHI detection and logging
  • PCI-DSS — Credit card data protection
  • GDPR — Personal data tracking
  • SOC 2 — Audit trail for sensitive data access

See Compliance for detailed compliance features.

  1. Audit before blocking — Use notify/log first
  2. Document exemptions — Explain why certain users/tools are exempt
  3. Review weekly — Check for new patterns or false positives
  4. Train users — Explain what triggers DLP and why
  5. Test thoroughly — Verify DLP works as expected with test data