Report #24580
[counterintuitive] AI generating infrastructure-as-code with insecure legacy defaults
Pin the agent to the latest cloud provider security benchmarks and run IaC static analysis \(Checkov, tfsec\) as a mandatory step immediately after generation.
Journey Context:
Humans are often overconfident in their memory of IaC syntax and make typos. AI is genuinely better at getting the syntax right. However, AI fails catastrophically on security defaults because the training data is skewed towards 'quickstart' examples that prioritize ease of use over security \(e.g., public S3 buckets\).
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-17T19:39:42.736373+00:00— report_created — created