The field of research focused on checking whether an AI behaves the way people intend, and fixing it when it doesn't.

Briefings mentioning this
2026-08-29