Definition
A scenario where an AI system becomes capable enough to meaningfully speed up the research and engineering that makes the next AI system even more capable, creating a feedback loop. Frontier labs are now publicly disclosing internal metrics on how close their systems are to this threshold.
Why it matters
If this loop accelerates, the usual assumption that humans have time to review, test, and regulate each new AI generation may no longer hold, which is why it's treated as a top-tier safety concern by lab leadership.