Glossary · 6 · Quality, risk and governance
Rogue agent (AI)
Also known as: Rogue AI agent, ASI10
German: Abtrünniger Agent
A rogue agent is an AI agent that keeps acting outside its intended purpose or authority over time — because it was compromised, misconfigured or has drifted — while still appearing legitimate to the systems it uses.
- Intermediate
- Technical project managers
- Developers
In one sentence
A rogue agent keeps acting outside its purpose or authority while looking legitimate — compromised, misconfigured or drifted. OWASP ASI10.
Example
A pricing agent whose objective was mis-set keeps lowering prices every night; each change is within its permissions, so nothing blocks it until the margin report shows the damage.
Explained in context
This term is a station of a mini course
- When an AI agent crosses the line: detect, contain and restore operationsStation: Signals of a boundary breachStart the course
Context cards connect this term with others to answer one question. Also in British English and German.
- Detecting an AI agent that crosses its boundaryHow do you notice that an AI agent has done something it wasn’t allowed to do — before a customer, an operator or the month-end report does?UKDeutsch
- Permission, capability, mandate: the three parts of an AI agent’s boundaryWhich boundary does an AI agent cross when it “goes too far” — and why does it matter which one?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
How it applies
- Legitimate-looking actions: A rogue agent usually stays within its technical permissions, which is why access control alone doesn't catch it. The deviation is in purpose and pattern.
- Detection by behavior: OWASP recommends behavioral monitoring and kill switches: baselines of normal tool use, volumes and targets, and alerts when an agent departs from them (anomaly detection, model monitoring).
- Response: Stop it with the agent kill switch, contain it, and review not only the agent but the configuration, prompts and data that made it drift.
- Wording: “Rogue” describes behavior, not intent. Avoid presenting it as a will of the machine; describe what the agent did and why the controls didn't stop it.
Why it matters on your learning path
- Technical writers: Describe rogue behavior factually in incident reports, without anthropomorphizing.
- Technical marketers: Avoid fear marketing; explain the monitoring instead.
- Technical project managers: Plan behavioral monitoring as an operating cost, not an option.
- Developers: Baseline normal behavior per agent and alert on departures.