AI is writing more of your Terraform
AI agents generate insecure Terraform at alarming rates (19–40% pass rates on real-world tasks per NeurIPS/FSE/ICSE benchmarks), failing in four predictable ways: permissive IAM defaults, omitted security blocks, hardcoded credentials, and stale provider attributes. These misconfigurations—wildcard Action: "*", missing storage_encrypted, plaintext secrets in Git—mirror patterns behind real cloud breaches. The article details each failure mode with concrete examples and explains why HCL is harder for LLMs than YAML/JSON, making detection at merge-gate critical as AI authorship scales.
This post is licensed under CC BY 4.0 by the author.