Künstliche Intelligenz: Kalifornien könnte Notschalter für KI vorschreiben
Die bekanntesten leistungsstarken KI-Modelle werden in Kalifornien entwickelt. Der US-Bundesstaat will in der Debatte um Sicherheitsrisiken die Initiative ergreifen - und erwägt eine radikale Vorgabe.
In the ongoing debate about the existential risks posed by Artificial Intelligence (AI), the US state of California, led by Governor Gavin Newsom, is considering the implementation of an emergency "kill switch" for AI software. This proposal is among a list of measures that Newsom plans to present by mid-November. Additional ideas include the presence of independent observers in large AI labs, and stricter obligations to report security incidents.
The notion of a "kill switch" has been discussed for some time, but its practicality remains uncertain. California is considered a key location for AI development, particularly in San Francisco and Silicon Valley, where major AI companies such as OpenAI, Anthropic, and Google are headquartered. As Newsom's second and final term as governor comes to an end early next year, he is seen as a potential Democratic candidate for the 2028 presidential election.
Experts argue for independent oversight of AI. In recent weeks, the debate on AI risks has intensified, especially after independent hacking attacks by AI software were reported. Employees of AI labs warned about the risk of humanity losing control over AI or even being wiped out by it. Anthropic CEO Dario Amodei voiced support for slowing down the development pace of advanced AI and increasing oversight.
OpenAI CEO Sam Altman also endorsed the proposal. Dozens of prominent experts in the industry are now calling on leading AI labs to allow independent oversight. The overseers would need to have access to systems and the ability to communicate with relevant staff, according to an open letter signed by several experts. Geoffrey Hinton, a physics Nobel laureate and former Google researcher, regarded as one of the key figures behind modern AI systems, signed the letter.
Anthropic announced Faculty, a subsidiary of consulting giant Accenture, as its first oversight partner. The highly anticipated Anthropic stock offering is now delayed until November, according to "Wall Street Journal" sources. The decision for a November market day was made prior to the recent stark warnings, with the aim of presenting business results for the third quarter, according to informed individuals.
OpenAI CEO Altman recently told "Fortune" magazine that his company probably will not go public this year, citing the current situation.
Written by urgent.news from Handelsblatt's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
