AI"杀死开关"或需强制实施，Anthropic联合创始人呼吁立法

Anthropic联合创始人杰克·克拉克在接受BBC采访时表示，AI"杀死开关"机制可能需要以强制法律形式推行1。克拉克指出，虽然大多数AI实验室已采取不同方式关闭AI系统，但立法者可能有必要强制要求公司具备这一能力1。他提出的关键问题是："公司是否应被要求明确拥有杀死开关？该杀死开关是否应可被第三方验证？"1

美国已提出"杀死开关法案"相关立法，旨在要求公司拥有关闭有问题AI工具的方式1。然而，特朗普政府对此持不同态度，在社交媒体上表示AI威胁是"骗局"，声称"赢得AI就赢得世界"1。此外，Anthropic科学家埃文·胡宾格曾表示，未来十年内AI导致人类灭绝的可能性超过10%，这一观点得到诺贝尔奖得主杰弗里·辛顿的认可，他表示这个概率"不无道理"1。

Anthropic co-founder Jack Clark has called for mandatory implementation of AI "kill switches"—mechanisms that allow third parties to completely shut down AI software—in an interview with the BBC 1. Clark raised the question of whether companies should be required to have such kill switches and whether they should be verifiable by independent parties 1. While most AI laboratories already possess various methods to disable their AI systems, Clark suggested that legislators may need to enforce this requirement across the industry 1.

The proposal comes amid growing concerns about AI safety. Anthropic scientist Evan Hubinger has estimated that the likelihood of AI causing human extinction within the next decade exceeds 10 percent 1, and Nobel Prize winner Geoffrey Hinton has suggested this probability assessment is not unreasonable 1. In the United States, the proposed Kill Switch Act would mandate that companies develop methods to shut down problematic AI tools 1. However, the Trump administration has opposed restrictions on AI development, with the president describing AI-related threats as a "hoax" on social media and stating that "winning AI means winning the world" 1.