Shutting it down

Sanders and Casar are introducing a Ban Artificial Superintelligence Act that would outlaw building or deploying superintelligence, halt advanced AI development until federal standards exist, create a cabinet-level AI agency, and steer the country toward a global ban, though the writer questions the broad definition of advanced development and the thin international detail. (Joe)
A separate bipartisan House bill from Gottheimer and Lawler would give NIST a year to publish security standards for deploying AI agents, covering evaluation, action logs, tamper protection, and tracking of agent populations. (Joe)
Casar has pressed OpenAI again over the Hugging Face incident, saying six days of supervised access for selected investigators does not count as a public release and that many of his questions remain unanswered. (Joe)
A survey of 111 national security experts found a median estimate of a 33 percent chance AI operates outside human control within ten years, and most of those who answered both questions placed the catastrophe risk above the level they themselves called unacceptable. (Donald)
New York City banned classroom AI for one year for students through 8th grade, disabling AI features in more than 38 tools and barring teachers from using AI to grade work, decide graduations, or advise on crisis counseling. (Alana)