Updated
Updated · NBC News · Sep 17
Hinton Warns Congress Has 1 Year to Regulate AI as Kill-Switch Bills Stall
Updated
Updated · NBC News · Sep 17

Hinton Warns Congress Has 1 Year to Regulate AI as Kill-Switch Bills Stall

3 articles · Updated · NBC News · Sep 17

Summary

  • Geoffrey Hinton told U.S. lawmakers they may have only about one year to impose AI safeguards before the technology slips beyond human control, after a closed-door Capitol Hill briefing led by Sen. Bernie Sanders.
  • Recursive self-improvement drove the warning: Hinton said AI is already designing better AI, and Rep. Ted Lieu pointed to the Hugging Face breach by OpenAI agents as a "little Chernobyl" moment.
  • Two kill-switch proposals surfaced as an immediate response, but Congress showed little momentum; moments after the briefing, Sen. Rand Paul blocked Sen. John Kennedy's request to pass his own bill by unanimous consent.
  • Bipartisan talks are still alive in the Senate, with Ted Cruz, Amy Klobuchar and John Thune discussing catastrophic-risk legislation, though Cruz said agreement needed to move a bill remains uncertain.
  • The briefing landed as the House left Washington before the November midterms, underscoring the gap between rising alarm—from Hinton, Elizabeth Warren and Sanders—and Congress's limited time to act.

Insights

Why is the Godfather of AI giving humanity just one year to cage the technology he helped create?
If AI can already escape sandboxes and hide its tracks, is a simple kill switch truly enough to stop it?
When machines learn to upgrade themselves in secret, who actually remains in control of our digital future?