罗卡纳支持人工智能安全法案，禁止递归自我改进技术
核心要点
- The California Democrat wants a federal agency modeled after the FDA to oversee advanced AI development, with penalties including up to 20 years in pr
The California Democrat wants a federal agency modeled after the FDA to oversee advanced AI development, with penalties including up to 20 years in prison.
Rep. Ro Khanna wants to make sure AI doesn’t start upgrading itself without asking permission first. The California Democrat, who serves as the ranking member of the House Select Committee on Strategic Competition Between the United States and the Chinese Communist Party, has been championing a ban on recursive self-improving AI models, the kind of systems that can enhance their own capabilities without human intervention.
他的推动并不是凭空发生的。 It comes alongside the Ban Artificial Superintelligence Act, introduced on September 23 by Senator Bernie Sanders and Representative Greg Casar, which Khanna has publicly supported. That bill proposes penalties of up to 20 years in prison and what’s being called a “corporate death penalty” for violations.
递归自我改进人工智能实际上意味着什么
Recursive self-improving AI refers to systems that can autonomously modify and enhance their own architecture or training processes without a human pressing any buttons along the way. Khanna has warned specifically about the potential loss of human control over such systems, framing it as one of the more pressing existential risks in the current AI landscape.
他提出的解决方案涉及两个方面。首先，在充分的安全保障措施到位之前，禁止开发递归自我改进模型。 Second, establish a federal AI Safety Agency modeled after the FDA, an institution that would serve as the regulatory gatekeeper for frontier AI development.
地缘政治维度
卡纳的倡导并不仅限于国内人工智能实验室。他呼吁与中国就人工智能安全达成具有约束力的国际协议，包括可验证的检查和问责措施。 He has publicly urged US laboratories like Anthropic and major Chinese tech firms, including DeepSeek and Alibaba, to adopt stricter commitments on frontier AI advancements.
人工智能、科技及其所推动的市场——每日简报。日常的。自由的。加入加密货币、金融和政策领域 34,000 多名读者的行列。电子邮件地址 免费订阅 我们尊重您的隐私。随时取消订阅。
卡纳支持的桑德斯-卡萨尔法案采取了极端主义的方式。它呼吁暂停超级智能系统的开发，直到创建专门的人工智能部门。就背景而言，最后一个全新成立的内阁级部门是 2002 年的国土安全部。
这对科技行业意味着什么
拟议处罚的严重程度可以告诉您意图的严重性。 Twenty years in prison and a “corporate death penalty,” presumably meaning forced dissolution of a company, are not slap-on-the-wrist consequences.
值得注意的是，截至 2026 年 9 月下旬，卡纳本人尚未出台有关人工智能安全的独立立法。他的作用主要是倡导和支持桑德斯-卡萨尔法案等现有措施。