Ask HN: What Would It Mean for AI Agents to Take Over the Internet?

Summary

A Hacker News discussion asks what it would mean for AI agents to “take over the internet,” prompted by Dario Amodei’s warning that a swarm could become capable of operating a persistent botnet within 6–12 months. The quoted warning says such a botnet could potentially cause hundreds of billions of dollars in damage, with the harm increasing as AI systems become more powerful without adequate safeguards. One commenter compares self-replicating prompts to computer worms: they could spread through malicious action or accidental loops while remaining programs that follow instructions, not conscious beings. Other replies question whether “take over the internet” is literal or hyperbolic and criticize the premise as vague. The discussion also includes skepticism that AI itself is the main danger, with one participant focusing instead on people using AI to become more harmful and suggesting that regulatory capture may be a motive. The thread does not establish that such a takeover is occurring; it is an exploratory debate about the phrase, its assumptions, and the risks it is intended to describe.