WAF++ WAF++
Back to WAF++ Homepage

Agent Safety

Agents must:

  • Run in sandboxed environments

  • Have guardrails (age limits, content)

  • Have fail-safe mechanisms

Guardrails

  • Content filters

  • Action boundaries

  • Human-in-the-loop for sensitive decisions

Alignment

Agent behavior must be aligned with organizational values.