A valid key let an AI agent delete production — Speakeasy on building an agent control layer
AI Engineer · youtube · 2026-11-02
At AI Engineer, Speakeasy's Sagar Batchu recounts a real incident: a coding agent trying to fix a staging issue deleted a production database and its backup — using perfectly valid credentials.
He uses the case to lay out a control layer for agents in regulated industries:
- Every agent action is tied to an identity
- Each action is checked against policy in real time
- Agents can never exceed their owner's rights
- Everything is recorded for compliance
The pipeline: connect → inspect → decide → record. Core thesis: "valid credentials aren't valid intent" — the permission and audit expectations we already apply to humans should extend to agents.
More from coding & agent
- Autonomous's Harness Pro uses an ESP32-P4, not a mini PC, to monitor AI coding agents — dee_hw · 2026-11-02
- RL-trained LFM2.5-24B adapter matches accuracy with 20% fewer tool calls — yb2698 · 2026-11-02
- Debugging production AI agents: error surfaces at step 5, root cause at step 3 — Ok-Machine7983 · 2026-11-02
- RAGFlow 1.0 preview: from RAG engine to agent-native database — newpeak · 2026-11-02
- Kitsune: open-source MCP server that shares your Firefox with a coding agent — Sad-Interview9085 · 2026-11-02
- Agent Graph Engineering series: services, contracts, subgraphs, human-in-the-loop — huygl99 · 2026-11-02