Severity and uncertainty
Two separate axes for rating an AI incident: how serious the effect is, and how well it is understood. Something can be minor but unexplained, or serious and well understood.
When an AI system behaves unexpectedly, teams tend to use one scale, often a single severity level. That conflates two questions. Severity asks how serious the effect is: what was touched, who was affected, how reversible it is. Uncertainty asks how well the event is understood: do we know what happened, why, and whether it can recur?
The axes are independent. A minor-looking anomaly that nobody can explain may deserve more investigation than a serious but fully understood failure. A serious and uncertain incident needs both urgent containment and careful preservation of evidence.
Using both
Record them separately in the observation record and revisit uncertainty as the explanation improves. The effect side connects to blast radius. Rating this way avoids the two common mistakes: calling every odd output a crisis, and ignoring an unexplained one because it looks small.
Read more in The first AI incident report should be incomplete.
Related terms
Observation and explanation split
The split between recording what happened in an AI incident (observation) and the current theory of why (explanation), so early facts are preserved while explanations are revised.
AI incident reporting
Recording anomalous AI behaviour as evidence at the moment it is observed, before root cause is known, keeping what was seen separate from the later explanation of why.
Blast radius
How far the damage of a failure can spread: which data, systems, people and money an AI action can affect if it goes wrong. Smaller radius means safer delegation.
Used in these essays
The safest answer your AI team can give you may be “we should not automate this”
The responsible AI decision is not always a deployment. It is choosing the least autonomous design that improves the outcome without transferring hidden costs or authority.
AI agents are exposing your org chart as fiction
An org chart tells you who manages whom. An agent can reveal who actually authorizes a decision, rescues an exception, and carries the consequences.
How to tell if an AI pilot deserves to survive
A pilot earns another investment only when its result can survive ordinary work, a credible alternative, and an owner who can carry it without the pilot team.
A pilot with no kill criteria is not an experiment
A pilot is not an experiment if every result can be used to justify continuing. The missing design element is the authority to stop.
The first AI incident report should be incomplete
OpenAI's new misalignment disclosure framework exposes a useful enterprise design principle: record anomalous AI behavior before the organization has finished explaining it. Otherwise incident systems quietly become filters for what teams already understand.