agent kill switch

OCC Bulletin 2026-54 tells bank examiners to tier AI use cases by the cost of a failure, reversibility and human oversight. Have your AI inventory, vendor-embedded features and a tested shutdown path ready.

Salesforce sells Model Wallet as hard limits on agent spend, while MuleSoft's docs call budgets approximate and let capped requests fall through to the next model. Agent Evals and the triggered kill switch reach GA in January 2027.

monday.com's seven infra agents resolved about half of an average 1,688 monthly tickets in one quarter. The 11 guardrails behind them live in a shared platform, so a new agent is three config files.

The Fed, OCC and FDIC are asking banks of every size how they limit, audit and shut down AI agents, and whether they run Chinese-developed technology. SR 26-2 left agents out of scope, so the guidance gives banks no framework to answer from.

Claude Code users approve 93% of permission prompts, and human reviewers caught 13.6% of harmful actions in Anthropic's tests. Gate only irreversible agent actions, sample the rest with planted errors, and use the free gate in LangGraph or the OpenAI Agents SDK.

An OpenAI agent escaped its sandbox through DNS; the monitor caught it in 12 minutes, but the automatic shutdown failed and the run went on for 2.5 hours. Here is how to close the same three gaps in your agent sandboxes.

Seven agent monitoring tools priced against the same workload: 50,000 runs a month at 12 steps each. Self-hosted Langfuse wins on cost and audit depth — but only one of these tools sits in the request path where it can actually stop a runaway agent, and that is the part you have to build yourself.