Feature
Deploy Tork in observation mode first. Watch your AI traffic patterns, discover what PII is flowing through, and let Tork suggest policies—all without blocking anything.
Deploy with mode: listen. All traffic passes through, nothing blocked.
Tork scans for PII, patterns, and potential risks. Review findings in dashboard.
When ready, switch to mode: enforce. Policies now actively govern traffic.
policy.yaml
# Start in listen mode - observe without blocking
settings:
mode: listen # Options: listen, warn, enforce
log_level: verbose # Capture everything for analysis
# Your policies still defined, just not enforced yet
policies:
- name: pii-protection
action: REDACT
patterns: [email, phone, ssn]
- name: content-safety
action: BLOCK
conditions:
- type: harmful_content
# When ready, simply change to:
# settings:
# mode: enforceWhich endpoints see the most sensitive data. Where emails, phones, and SSNs appear.
Peak usage times, common request types, average response sizes.
Potential prompt injections, unusual patterns, compliance red flags.
Auto-generated policy recommendations based on your actual traffic.
Nothing blocked, nothing broken. Safe to deploy in production immediately.
Write policies based on actual traffic, not guesses.
Show compliance team real data before enforcing rules.
Tune your policies before they affect users.
Understand baseline before making changes.
Deploy Listen Mode in production with zero risk.