Skip to content
BENIBIX Learn

AI Incident

Business💻 Information SystemsAI Adoption

Also known as Serious AI Incident, AI Incident Reporting

When an AI system does real harm. For example, a wrong answer that someone acted on.

Think of it like

It is like a report after a bike crash. You write down what went wrong so it does not happen again.

Example

In 2024 a tribunal in Canada made Air Canada pay a customer. Its chatbot had told him he could claim money back. The airline's own rules said no.

Why it matters

Every AI system gets things wrong sometimes. Firms that log and learn from incidents stop the same harm happening twice.

Definition

An AI incident is an event where an AI system causes real harm. It still counts when the AI only helps cause it. The harm might be to a person's health, their rights, their property or a public service. When harm could have happened but did not, it is often called an AI hazard. Firms log each incident, fix the cause and learn from it. Some laws make firms report the serious ones.

Common mistake

Treating it as just an IT outage. The harm can happen while every system is up and running.

Key takeaways

  1. 1
    Real harm. Harm happened, not just could have.
  2. 2
    Log and learn. Record it and fix the cause.
  3. 3
    Report serious ones. Some laws make firms report them.

Compare