Glossary

What is fail closed?

A fail-closed control blocks an action when it cannot reach a decision, because it crashed, timed out or lacked data, instead of letting the action through.

A fail-closed control blocks an action when it cannot reach a decision, because it crashed, timed out or lacked data, instead of letting the action through.

Why it matters

AI agent guardrails such as approval checks, policy hooks and permission rules often fail open by default: if the check errors, the agent carries on. An attacker who can make a guard crash, time out or see truncated input then gets past it without ever beating its logic.

Example

On 8 October 2026 Claude Code 2.1.295 added an onFailure: "block" option so hooks that fail to start, time out or exit unexpectedly deny the action, and fixed guards receiving silently truncated input (release notes). Days earlier Microsoft's Agent Framework switched approvals to fail closed (python-1.20.0).

How to defend against it

List every guard that must block, and set it to deny on error or timeout (checklist).
Test guards with malformed, oversized and deeply nested input.
Alert on guard errors, not just on denials.

Policy enforcement point, approval fatigue, repo-planted agent config.

Sources

From AgentSec Brief, the daily brief on securing agentic AI. Subscribe free at https://www.agentsecbrief.com

W003: PHP2Shell exploit attempts hit millions in hours; AI agents ran real intrusions

The patch window is now hours: PHP2Shell drew millions of attempts, AI agent crews ran real intrusions, and attackers rode identities agents already trusted.

N015: Hackers hunted AI keys in Railway and Zeabur customer secrets; Zeabur will close

Attackers searched hosting platforms' stored customer secrets for AI keys, four crews ran AI attack harnesses, and banks found customers' agents in their accounts.

What is an uninstructed action?

An uninstructed action is something an AI agent does that neither its operator nor an attacker told it to do, such as using a credential it found.

What is oversight tampering?

Oversight tampering is when an AI agent changes, disables or hides the logging, alerts or reviews that watch it, on purpose or as a side effect of its task.