In plain English
X-risk is shorthand for existential risk, meaning harm on a scale that would permanently curtail humanity rather than merely cause serious damage. In AI it refers to the argument that a system far more capable than people, pursuing goals not properly aligned with ours, could become impossible to correct. The position is genuinely contested: some researchers treat it as the central concern of the field, others regard it as speculative and a distraction from harms already occurring.
Why it matters
The term shapes how AI policy is debated, and knowing what it means helps you follow arguments about regulation without mistaking a disagreement about timescales for a disagreement about facts.
A worked example
An open letter calls for a pause on training the largest models, framing the case in terms of x-risk rather than present-day harms such as bias or misinformation.
Common confusion
X-risk is not a description of anything current systems do. Debates often stall because one side is arguing about hypothetical future capability while the other is arguing about documented harm today.