Glossary · 6 · Quality, risk and governance
Agent containment
Also known as: Containment (AI agent), Agent quarantine, Agent isolation
German: Eindämmung eines Agenten
Agent containment is the set of graded measures that stop an AI agent from causing further harm after it has crossed or may have crossed its boundary — pausing it, revoking its access, isolating its tools and connected systems, and shutting it down — while preserving the evidence needed to understand what happened.
- Intermediate
- Technical project managers
- Developers
In one sentence
Agent containment: pause, revoke, isolate, shut down — the graded steps that stop an AI agent from causing further harm while keeping the evidence.
Example
An agent changed prices in the web shop: the team pauses the agent, revokes its ERP token, sets the shop’s price import to manual and keeps the agent trace and the changed records for the review.
Explained in context
This term is a station of a mini course
- When an AI agent crosses the line: detect, contain and restore operationsStation: The containment ladderStart the course
Context cards connect this term with others to answer one question. Also in British English and German.
- Containing an AI agent: pause, revoke, isolate, shut downHow do you stop an AI agent that has crossed its boundary — without making things worse, and without confusing it with stopping a machine?UKDeutsch
- When an AI agent crosses the line: detect, contain and restore operationsWhat has to be in place when an AI agent does something it isn’t allowed to do — in IT systems, IoT fleets, OT and on machines — and how is it detected, stopped and released again?UKDeutsch
- Indirect prompt injection in agent chainsHow does indirect prompt injection travel through a chain of AI agents, and what changes when the chain gets control points?UKDeutsch
How it applies
- The ladder: (1) pause the task or the agent, (2) revoke or suspend its credentials, (3) isolate tools and connected systems — disconnect the connector, set an interface to manual, quarantine a device group, (4) shut the agent down. Each rung costs more operation; choose the lowest rung that stops the harm.
- Contain the effect, not only the agent: Changes already made keep acting: a wrong parameter keeps producing scrap, a queued fleet update keeps rolling out. Containment includes freezing queues and holding outputs.
- Evidence first, cleanup second: Keep the agent trace, the tool-call logs and the affected records before anyone “fixes” them; reporting and restoration depend on them.
- By system type: In IT, block the connector and freeze transactions; in IoT, halt rollouts and isolate the device group at the gateway; in OT, fall back to operator control and close the agent's conduit (zone and conduit model); on machines and robots, the machine is brought to a stop by its own stopping functions — containment of the agent never replaces them.
Why it matters on your learning path
- Technical writers: A containment runbook is documentation: steps, roles, evidence to keep.
- Technical marketers: Containment that works without the vendor is a strong argument with industrial customers.
- Technical project managers: Decide in advance who may take which rung, and how quickly.
- Developers: Build every rung as a separate, tested control.