Full Breakdown
AI Threats to Internet Control: Expert Views on a Possible “Swarm”
By Drooid · · How we work
Core Event
On September 28, 2026, CBS News reported that Anthropic CEO Dario Amodei warned of a potential AI “swarm” that could seize control of the internet. The article examined whether such a scenario is technically feasible, citing a range of security researchers who argue that the internet’s decentralized architecture and limited computing resources make a full-scale takeover unlikely.
Technical and Resource Constraints
Vishal Misra, a computing and artificial-intelligence expert at Columbia University, explained that the internet is a “network of networks” built to survive loss of any single component, with billions of devices linked through diverse routers, switches and operating systems. This heterogeneity, he said, acts as a defense because there is no central switch to flip.
John Thickstun of Cornell University noted that AI agents would need massive computational power, which depends on advanced semiconductors that are scarce and subject to export controls. Kevin Hamlen, executive director of the Cybersecurity Research and Education Institute at the University of Texas at Dallas, added that criminals would require sustained access to dedicated computing resources before posing a serious danger.
Expert Perspectives on Potential Impacts
Moshe Vardi of Rice University cautioned that AI does not need to orchestrate a global catastrophe; localized disruptions—such as hacking power stations or issuing false commands—could cause industrial disasters. Roman Yampolskiy highlighted that AI agents could coordinate attacks on financial markets, hospitals and utilities around the clock, a capability beyond human hackers.
Anthony Aguirre, co-founder of the Future of Life Institute, likened rogue AI agents to an invasive species that could displace existing cyber-ecosystems, suggesting a gradual erosion of security rather than a single, dramatic takeover.
Verbatim Quotes
- “There would be plenty to do in trying, say, to keep one's intelligence up to the standard set by the machines, for it seems probable that once the machine thinking method had started, it would not take long to outstrip our feeble powers,” — Alan Turing
- “The internet is a network of networks that billions of devices connect to through routers, switches and Wi-Fi access points,” — Vishal Misra, a computing and artificial intelligence expert of Columbia University's engineering school
- “AI agents operating in cyberspace would have to somehow obtain that,” — John Thickstun, a computer scientist and AI learning expert at Cornell University
- “It might also impersonate humans and issue false commands. A bad actor might use it that way deliberately,” — News. Gary Marcus
- “My big worry is not about AI taking over the whole internet, but about some industrial disaster that's caused by an out-of-control AI,” — Moshe Vardi, a computer scientist at Rice University