Particle.news
Download on the App Store

U.S. House Proposes Bipartisan Law Requiring Built‑In AI Shutdown Controls

The measure would give the Homeland Security Secretary authority to order staged interventions for dangerous systems and impose steep fines, testing how to protect infrastructure while preserving U.S. tech leadership.

Overview

  • The bill would force AI developers to build layered shutdown tools that can stop inference, throttle computation, block users, or fully power down models and would let the Homeland Security Secretary order those actions for serious threats.
  • Sponsors wrote specific intervention triggers into the text, including cases where an AI resists shutdown commands or causes major harm measured by death tolls or large economic losses, and proposed fines up to $20 million per day for noncompliance.
  • Security concerns about advanced AI enabling cyberattacks on banks and critical infrastructure and a high‑profile industry alert increased urgency for lawmakers and helped prompt the proposal.
  • President Trump said he will raise AI at the September U.S.–China summit and meet industry leaders next week to push for regulation that protects security without undermining U.S. industrial leadership.
  • The bill remains a developing proposal and faces practical and legal questions over how a federal shutdown order would be enforced across complex AI systems, so its passage and implementation timeline are still uncertain.