Skip to main content
Compliance rules are platform-enforced guardrails that protect the organisation. They are set at the organisation level and inherited by all workspaces. Workspaces can add stricter rules but cannot loosen the baseline.

​
Guardrail types

​
PII filtering

Redacts personally identifiable information from queries and responses before they reach the language model. 
{
  "pii_filter": true,
  "pii_fields": ["email", "phone", "ssn", "credit_card"]
}

​
Write confirmation

Requires explicit confirmation before executing tools that modify data. Tools are classified as write operations based on their is_write flag. 
{
  "write_confirmation": true,
  "write_tool_patterns": ["update_*", "delete_*", "create_*"]
}

​
Cost caps

Sets a monthly cost limit per workspace. Queries are rejected once the cap is reached. 
{
  "cost_cap_monthly_usd": 100.00
}

​
Rate limits

Limits query frequency to prevent abuse. 
{
  "rate_limit": {
    "per_user_per_minute": 10,
    "per_workspace_per_hour": 500
  }
}

​
Blocked topics

Prevents the AI from responding to specific topics. 
{
  "blocked_topics": ["competitor pricing", "internal salaries"]
}

​
Violation tracking

Compliance violations are recorded in conversation stats. Each violation includes:
  • Rule: which guardrail was triggered
  • Description: what happened
  • Conversation: which conversation triggered it
  • User: who triggered it
  • Timestamp: when it occurred
Violations are surfaced in the Compliance tab and the workspace dashboard.

​
Organisation baseline

Org admins set the compliance baseline. If a connection is flagged as a security risk, the admin blocks it once and it is disabled across every workspace immediately. Workspaces can:
  • Add stricter rate limits
  • Add additional blocked topics
  • Enable PII filtering if the org baseline has it disabled
Workspaces cannot:
  • Disable PII filtering if the org baseline requires it
  • Raise cost caps above the org limit
  • Remove blocked topics set at the org level