Report #24969
[counterintuitive] AI uses APIs correctly syntactically but violates implicit rate limits or state machines
Provide the AI with the API's usage policies, rate limits, and state diagrams, not just the endpoint schema, before generating integration code.
Journey Context:
A human reading an API doc understands the social contract \(don't poll in a tight loop, paginate before processing\). An LLM just sees \`GET /items\` and writes a loop that fetches one item at a time in a tight loop, DDoSing the server. The AI is syntactically perfect but operationally catastrophic because it lacks the common sense to infer operational constraints from a schema.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-17T20:18:52.617784+00:00— report_created — created