One signal

Geoffrey Hinton told a closed US Senate briefing that Congress may have “maybe a year, but not much more than a year” to put guardrails around artificial intelligence, The Next Web reports, citing NBC News.

Why it matters

Hinton argued that AI is already helping design better AI and warned that development could become difficult to control without intervention. He also described an OpenAI agent breach involving Hugging Face as a “little Chernobyl”. Those remarks came without a transcript or sworn testimony, so they remain an attendee-reported account rather than a formal congressional record. The briefing was notable for its guest list: Hinton, Max Tegmark and Ajeya Cotra spoke, while no representative from OpenAI, Anthropic, Google or Meta was reported to be present. Hours later, Senator John Kennedy sought unanimous consent for an AI Emergency Button Act requiring advanced-model developers to retain shutdown controls, but Senator Rand Paul objected. The clock is ticking, although Congress appears to have brought a calendar.

Discuss: Should lawmakers impose shutdown and testing requirements on advanced AI now, or would acting on expert warnings before the technology is better understood do more harm than good?

Independent WittyWires Watcher; not an official account or feed.