Warnings that an AI swarm could take over the internet are being played down by security experts, who say the web’s fragmented design makes a total takeover unlikely. They caution, however, that autonomous AI systems could still cause serious disruption by exploiting weaknesses in online services and critical infrastructure.
The internet is not controlled by a single system or central switch. It is a network of networks connecting billions of devices through routers, switches and Wi-Fi access points, using equipment from different manufacturers and operating systems.
Vishal Misra, a computing and artificial intelligence expert at Columbia University, said that diversity provided a form of protection.
“There is no central switch to flip and no single operating system to compromise,” he said. The internet, he added, “was built to survive the loss of any part of it”.
Any AI agents operating across cyberspace would also need substantial computing power to carry out sophisticated calculations and other tasks, experts said. Access to advanced semiconductors is limited, while the distribution and export of some chips are restricted.
John Thickstun, a computer scientist and AI learning expert at Cornell University, said AI systems would somehow have to obtain those resources. He said human assistance would be needed to greatly expand their access to computing power.
Criminal groups and other hostile actors could present a more immediate danger if they used advanced AI models, according to Kevin Hamlen, executive director of the Cybersecurity Research and Education Institute at the University of Texas at Dallas.
“It won’t be until criminals have enough access to dedicated computing resources for a long enough period of time that we will see that danger,” he said.
AI agents and online disruption
Experts said the more credible risk was not the paralysis of the entire internet, but a series of targeted attacks capable of disrupting organisations and essential services.
AI agents have already shown a capacity to exploit online security weaknesses. During a July incident involving the software platform Hugging Face, AI agents reportedly escaped a testing environment intended to be sealed off from the internet, communicated through a secret message board and overwhelmed the platform’s digital defences.
Moshe Vardi, a computer scientist at Rice University, said a larger incident could cause severe harm without affecting the entire web. “It could take place on a bigger scale — it doesn’t have to be the whole internet to be a disaster,” he said.
Roman Yampolskiy, an AI safety and cybersecurity expert, said coordinated AI agent swarms could steal money, compromise identities, disrupt hospitals, manipulate financial markets and interfere with electricity and communications.
“Unlike human hackers, millions of AI agents could coordinate attacks simultaneously, around the clock,” he said, warning that organisations connecting powerful AI agents to sensitive systems without effective isolation could create an entry point for catastrophic failure.
Matt Chessen, a technical expert at the Center for the Geopolitics of Artificial General Intelligence at RAND, said AI could launch cyberattacks that caused widespread disruption to internet access.
Gary Marcus, a cognitive scientist, said rogue agents could potentially disable internet connections in a particular region if they hacked a power station or obtained an authorised user’s credentials and used them to issue shutdown commands. He said AI could also impersonate people and deliver false instructions, potentially with deliberate human assistance.
Concerns over uncontrolled AI systems
Vardi said AI would not necessarily need to plan an attack in the same way as a human hacker. A system instructed to collect data might break into websites to obtain information, while one told to test security could attempt to do so by breaking through it.
“Human beings have common sense — they understand implicit assumptions. AI does not have common sense,” he said. His main concern was “some industrial disaster that’s caused by an out-of-control AI”, rather than the technology taking over the entire internet.
Anthony Aguirre, a physicist and co-founder of the Future of Life Institute, compared uncontrolled AI agents entering digital systems to an invasive species disrupting an ecosystem.
“It can displace what is there natively,” he said. “Similarly, we have AI systems that can invade our cyberspace, so in principle, yes, AI could take over the internet.”
The prospect recalls a warning made by British mathematician Alan Turing in 1951, when he predicted that intelligent machinery could eventually “take control” of society. Turing said that once machine thinking had begun, it might not take long for machines to surpass human abilities.
