Your agent issued a bad refund — who's accountable? A production governance framework
No-Conflict4823 · reddit · 2026-09-30
A Reddit long-form post (by a founder building in this space) tackles accountability when AI agents act: bad refunds, account changes, deleted data.
Key arguments: a prompt is a request, not a permission — "never refund over $100" means little if the agent's credentials allow a $10,000 refund; buying a model doesn't make the vendor responsible for everything you let agents do. For consequential actions the author wants five controls: enforced limits checked at execution, a named approver, least-privilege access with secrets kept out of model context, safe retries (timeouts must not become duplicate payments), and independent, tamper-proof records of authorization and execution. Controls only cover actions that pass through them, so side channels remain a hole. Governance, the author argues, is what makes more autonomy possible — not human approval on every click. Open question: who owns agent authority on your team — engineering, security, or the business owner?
More from coding & agent
- Dimillian asks for a ChatGPT app-plugin orchestrator/board system, builder shares hooks approach — Dimillian · 2026-09-30
- ModRetro and OpenAI bring Codex-built games to the Chromatic handheld — Dimillian · 2026-09-30
- Kite ML launches RL Studio: teach robots new skills with one API call and a prompt — Ronangmi · 2026-09-30
- Indie Dev Ships an AI RSS Plugin for the DeepSeek Harness Ecosystem — vista8 · 2026-09-30
- First DeepSeek Harness plugin ships: Qiaomu AI RSS brings AI news and podcasts into your coding agent — vista8 · 2026-09-30
- DoorDash launches corporate ordering MCP and CLI so AI agents can place orders, Vercel and Cognition early adopters — Scobleizer · 2026-09-30