{
  "id": 8308920,
  "title": "Is an AI kill switch a real thing?",
  "url": "https://urgent.news/2026/09/18/is-an-ai-kill-switch-a-real-thing",
  "topic": "ai",
  "section": "AI",
  "published": "2026-09-18T20:31:33.000Z",
  "source": {
    "name": "Mashable",
    "slug": "mashable",
    "url": "https://mashable.com/tech/ai-kill-switch-newsom"
  },
  "original_language": "en",
  "account": "California Governor Gavin Newsom has intensified his state's efforts to regulate artificial intelligence (AI) with the issuance of an executive order on September 18. The order calls for the establishment of an expert-led panel that will strengthen the state's AI safety laws, including accelerating third-party safety evaluations and mandatory risk reporting. Additionally, the order includes the exploration of an emergency shut-off mechanism for frontier models, often referred to as an \"AI kill switch\". Newsom explained that the term encompasses a multitude of interpretations, depending on the context.\n\nThe concept of a universal \"off\" button for increasingly powerful AI agents has gained traction in legislative discussions and online debates. While the specific implementation of the kill switch remains a topic of debate, recent developments have seen two congressional representatives introduce the bipartisan AI Kill Switch Act. This bill proposes granting the Department of Homeland Security (DHS) the authority to slow down or halt AI systems deemed capable of causing catastrophic harm. It also mandates developers to implement mechanisms to throttle, suspend, or fully shut down covered AI systems.\n\nDespite the lack of legal requirements, individual AI labs have their own means to halt rogue systems. However, there are proposals for the development of kill switches that involve external safety auditors and industry safety standards that would mandate emergency shut-offs, governed by the federal government. Some doomsdayers argue for a sweeping fail-safe for all AI, with experts questioning the effectiveness of binary safety regulations. AI companies can easily manipulate these terms, and such measures may become outdated as the technology continues to evolve.\n\nThe complexity of implementing an AI kill switch becomes even more apparent when considering the nature of AI systems. Unlike traditional physical environments, AI systems are not confined to a single location. Rogue AI agents may need to be hunted down on the open internet to pinpoint and contain their operators. This process may require a coordinated global response to isolate, stop, and degrade the AI to a manageable state. Moreover, questions of authority arise, as it is unclear who should have the power to make the call. Is it the original developer, an industry process, or an agency like Homeland Security? Despite these challenges, policymakers and industry observers are urging technologists to consider the possibility of an AI kill switch as a potential solution to address concerns about AI becoming sentient and maliciously determined to harm humanity.",
  "summary": "California Gov. Gavin Newsom just issued an executive order convening AI experts to explore an AI kill switch. Others are less certain it's worth our time.",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}