Definition
Automated controls in the AI deployment path that enforce budget caps, detect runaway token consumption, and prevent unplanned LLM spend — shifting cost governance from monthly review to build-time enforcement.
Key Points
- Triggered at amazon after multi-million token overruns (FT-attributed Jul 2026) (2026-07-30-amazon-ai-overspend-headlinesbriefing)
- Complements anti-gamification (shutting tokenmaxxing leaderboards on kiro)
- Industry signal for FinOps of coding-agents and enterprise LLM apps