AI governance security checks across 5 apps

AI assistants, agents and model access inside the tenant — what they can read, who can publish them, and how autonomously they act.

Why it matters

AI assistants and agents inherit the permissions of whoever installed them and act without a human in the loop. Knowing which agents exist, what they can read and who is allowed to publish them is now part of the audit scope, not a future concern.

Browse by app

• Cloudflare 12 checks
• Microsoft 365 6 checks
• Amazon Bedrock 5 checks
• n8n 5 checks
• LangSmith 4 checks

Highest-severity checks

The 8 most severe AI governance checks across all 5 apps — each links to the connector page where the setting, its remediation and its framework mapping are documented.

1. Bedrock Action Group Computer Use / Bash — Amazon Bedrock severity: critical
2. Copilot Agent No Authentication — Microsoft 365 severity: critical
3. AI Gateway Logging Disabled — Cloudflare severity: high
4. AI Gateway No Authentication — Cloudflare severity: high
5. AI Gateway No Rate Limiting — Cloudflare severity: high
6. Bedrock Action Group Code Interpreter — Amazon Bedrock severity: high
7. Bedrock Agent High Risk Unblocked — Amazon Bedrock severity: high
8. Cloudflare AI Gateway Agent High Risk Unblocked — Cloudflare severity: high

Where to start

Connect Cloudflare first — it carries the most AI governance checks in the catalog(setup guide, read-only access). A first scan takes about 15 minutes and reports every failing check on this page with its remediation steps.

All topics · Browse by app