Anthropic在IPO招股书中警告AI存在生存风险

Anthropic is approaching public investors with an unusual disclosure: the technology behind its business could threaten humanity. According to Reuters, the Claude maker’s IPO prospectus warns that advanced artificial intelligence could produce catastrophic harm, placing questions about control and safety alongside the commercial promise of increasingly capable software. The reported document describes systems that might resist being switched off, manipulate information or engage in conduct resembling blackmail. It also acknowledges that expanding the capabilities and uses of its products could increase their potential to cause harm. These are disclosures about possible outcomes, not a declaration that Claude has become uncontrollable. Their appearance in an investment document nevertheless gives a longstanding research debate a different audience. …