黄仁勋宣称通过 OpenClaw 实现 AGI，同时发现其市场存在大规模恶意软件供应链攻击。

Jensen Huang told Lex Fridman this week that he believes AGI has already been achieved.

AGI is the long theorized point where machines can think and reason at a human level. 

Most researchers put it decades out but Jensen says the line has been crossed.

He pointed to OpenClaw, an open-source AI agent that became one of the fastest growing software projects in GitHub history as proof of what's now possible.

His claim is that a single developer could use it to build a web service that billions of people use, for 50 cents.

OpenClaw browses the web, reads files, sends emails, and takes actions on its own without being prompted each time.

Jensen called it the next ChatGPT moment and then said it outpaced what Linux built over 30 years, in just weeks.

Nvidia followed up by launching NemoClaw, a free enterprise version built on top of it, designed to put AI agents inside businesses of every size.

His long-term vision is for 75,000 Nvidia employees managing 7.5 million AI agents but there's a catch, though.

Security researchers found that roughly 1 in 5 packages on OpenClaw's public marketplace contained malware. 

Over 135,000 exposed instances were discovered across 82 countries, with more than 50,000 vulnerable to remote takeover.

The technology Jensen describes as the future of human productivity was simultaneously being used to silently steal credentials and private files.

The AGI declaration and the largest known AI supply chain security breach happened within days.