Prompt Inspection
Barracuda SecureEdge Prompt Inspection helps organizations monitor and control interactions with generative AI services. It analyzes prompts submitted to AI applications, allowing security teams to apply data protection and content safety controls to AI traffic in the same way they manage web and email traffic. Using Prompt Inspection, you can detect sensitive information, enforce security policies, and audit AI activity across your organization. AI Prompt Inspection is available with the SecureEdge Premium Access plan.
Key Features
Policy Templates – Pre-built options (e.g., Block Malicious Prompts, Monitor All AI Activity, Protect Sensitive Data in AI, Safe Use for Schools, Professional Advice Guardrails, Brand and Workplace Conduct, Monitor Consumer AI Web Chat, Allow Trusted Enterprise AI, and Lockdown — High Security) enable quick setup without custom policy creation.
Flexible Targeting – Apply policies to specific users/groups, IP networks, all AI tools, or individual AI applications.
Classifiers – Enable or disable classifiers such as Prompt Injection, Gibberish, and Toxic Content to control which classifiers are used when scanning prompts in Scan & Log and Scan & Enforce policies. You can also define custom banned keywords based on your organization’s security requirements.
Policy Priority – Policies evaluate top-to-bottom; the first match determines the action. A default applies if none match.
Scan Actions – Choose Scan & Log (log without blocking) or Scan & Enforce (block non-compliant prompts).
Prompt Auditing – Logs all detected events for compliance and review.
Global Defaults – Set organization-wide Strict Mode, inline timeout, and prompt word limits (overridable by policies).
For more information, click on the topic you are interested in.