Story 5: NIST, the NSA, and CISA published ten voluntary practices for labs training AI agents.
They define a new kind of incident, an agent taking significant actions nobody told it to take, and ask labs to stop such runs automatically and report incidents to the public.