Shadow AI is the unapproved use of AI tools or services outside governed, auditable processes. Its most dangerous form is invisible production usage: code serves traffic while nobody knows which parts were generated or how. section 10.5
Banning desirable tools creates a prohibition trap. Developers may continue using them through endpoints and services that evade security review, auditing, and operational visibility. Provision beats prohibition: make the sanctioned path faster, safer, and easier to use. section 10.5
A practical response combines:
- Sanctioned AI tools with proper audit trails. section 10.5
- Visible engineering checks, including lint-as-architecture and triage. section 10.5
- A deliberate effort to improve the governed path rather than merely punish the shadow path. section 10.5

