SealGate

Understanding Security Flags

Learn about security flags and the Lethal Trifecta.

SealGate uses three flags to track risk in AI sessions.

Security Flags

📘 Private Data Access

Triggered by: Reading files, querying databases, or accessing internal documents. Risk: Sensitive information could be exposed.

🌐 Untrusted Content Exposure

Triggered by: Fetching web pages or calling external APIs. Risk: The AI could receive malicious instructions (prompt injection).

✉️ External Communication

Triggered by: Sending emails, posting to Slack, or making external API calls. Risk: Confidential data could be exfiltrated.

The Lethal Trifecta

The "Lethal Trifecta" occurs when a session has all three flags active:

StatusMeaning
✓ Private DataAI has seen confidential info.
✓ Untrusted ContentAI may have received malicious instructions.
⏳ External CommunicationAI is attempting to send data externally.

Protection: SealGate automatically pauses any action that completes this trifecta, requiring your manual approval to proceed.

Viewing Flags in the Dashboard

The Sessions view uses colored dots to show active flags:

  • 🔵 Blue: Private Data Access
  • 🟡 Amber: Untrusted Content Exposure
  • 🔴 Red: External Communication
Security flags in sessions table

Risk Levels

  • Low (green): 0 flags
  • Medium (amber): 1 flag
  • High (red): 2+ flags

ACL Levels

Access Control Levels (ACL) provide additional protection:

LevelMeaning
PUBLICNon-sensitive data.
PRIVATEInternal/confidential data.
SECRETHighly sensitive data.

Enforcement: SealGate automatically blocks high-to-low data flows (e.g., reading SECRET data then posting to a PUBLIC channel). These blocks do not ask for approval - they are prevented by default.


For Admins: You can classify tools and set ACL levels from the Access Control page and via Policy Rules.