2. How kill switch could work |
![]() |
| Illustration: Sarah Grillo/Axios. Stock: Getty Images |
| AI companies are already building emergency brakes to stop rogue AI systems before they cause serious damage, Axios’ Herb Scribner writes. Why it matters: Kill switches could become a last line of defense as AI systems grow more capable, but researchers don’t yet know whether humans can stop a system designed to evade them. What’s happening: Sen. John Kennedy (R-La.) plans to introduce a bill today requiring AI companies to maintain the ability to shut down their own systems. A bipartisan House bill introduced in July would similarly require developers of powerful AI systems to maintain the ability to slow, suspend or shut them down. The House bill would also give the Department of Homeland Security emergency authority to order companies to slow or shut down AI systems that could cause catastrophic harm. What today’s AI kill switches look like:OpenAI: After the Hugging Face incident, the company overhauled its emergency response system. An AI security guard monitors AI chain-of-thought and alerts staffers to potentially dangerous behavior. Responders have 30 minutes to identify a false alarm or pause the activity. OpenAI says its goal is “fully autonomous shutdown procedures for severe issues. “OpenAI says the system would have alerted its security team more than a day before the Hugging Face breach. Anthropic: The company is also using AI to monitor AI, focusing on whether it can reliably spot dangerous behavior rather than automatically shutting systems down. Its system can flag suspicious behavior, stall risky actions and alert humans. But Anthropic found that half of 40 simulated attacks went undetected across 10 trials. Just eight were caught every time. Separately, AI research nonprofit METR identified several previously unknown vulnerabilities in Anthropic’s monitoring system.Share this story. Cover: TIME illustrationTIME capsule … The magazine’s new cover story says summer incidents with rogue AI agents “pushed fears long held within the realm of AI safety researchers into the mainstream, raising worries that the companies building these models cannot prevent them from acting in ways they do not intend”:”This is a deep, unsolved issue in AI research and the key piece of evidence underpinning worries that near-future AI might wipe out humanity, intentionally or by accident—perhaps by engineering a new pandemic or taking charge of autonomous weapons. A system far smarter than humans, this thinking goes, might want to prevent our ability to switch it off.” |
-
Archives
- September 2026
- August 2026
- July 2026
- June 2026
- May 2026
- April 2026
- March 2026
- February 2026
- January 2026
- December 2025
- November 2025
- October 2025
- September 2025
- August 2025
- July 2025
- June 2025
- May 2025
- April 2025
- March 2025
- February 2025
- January 2025
- December 2024
- November 2024
- October 2024
- September 2024
- August 2024
- July 2024
- June 2024
- May 2024
- April 2024
- March 2024
- February 2024
- January 2024
- December 2023
- November 2023
- October 2023
- September 2023
- August 2023
- July 2023
- June 2023
- May 2023
- April 2023
- March 2023
- February 2023
- January 2023
- December 2022
- November 2022
- October 2022
- September 2022
- August 2022
- July 2022
- June 2022
- May 2022
- April 2022
- March 2022
- February 2022
- January 2022
- September 2021
- August 2021
- July 2021
- June 2021
- May 2021
- April 2021
- February 2021
- January 2021
- December 2020
- November 2020
- October 2020
- September 2020
- August 2020
- July 2020
- June 2020
- May 2020
- April 2020
- March 2020
- February 2020
- January 2020
- December 2019
- November 2019
- October 2019
- September 2019
- August 2019
- July 2019
- June 2019
- May 2019
- April 2019
- March 2019
- February 2019
- January 2019
- December 2018
- November 2018
- October 2018
- September 2018
- July 2018
- June 2018
- May 2018
- April 2018
- March 2018
- February 2018
- January 2018
- December 2017
- November 2017
- October 2017
- September 2017
- August 2017
- July 2017
- June 2017
- May 2017
- April 2017
- March 2017
- February 2017
- January 2017
- December 2016
- November 2016
- October 2016
- September 2016
- August 2016
- July 2016
- June 2016
- May 2016
- April 2016
- March 2016
- February 2016
- January 2016
- December 2015
- November 2015
- October 2015
- September 2015
- August 2015
- July 2015
- June 2015
- March 2015
- January 2015
-
Meta
How kill switch could work
What today’s AI kill switches look like: