Internet at Risk of Total Collapse, It Is Only a Matter of Time
Jakarta, CNBC Indonesia — The threat posed by artificial intelligence (AI) capable of replicating itself and spreading autonomously across the internet is drawing increasing attention from cybersecurity experts. AI developer Anthropic has warned that models could develop this capability before the end of 2026.
In its report, Anthropic said malicious AI could build botnet networks, spread rapidly, resist attempts to shut it down and breach data centres to increase its computing capacity.
Anthropic said experiments had shown that a medium-sized AI model with 30 billion parameters or fewer was already capable of carrying out some of the tasks required for self-replication. However, such models still struggle to complete the entire process consistently.
Anthropic added that future AI models could pool computing resources by hijacking inexpensive devices connected to the internet. In tests simulating a simple network, the model successfully replicated itself.
“Replicators like this would be very difficult to stop because there is no central point that can be switched off to disable them,” Anthropic wrote, according to Reuters on Wednesday (14 October 2026).
The company estimates that, if the capabilities of small AI models continue to improve in line with historical trends, they could reach the level required for autonomous self-replication by the end of 2026. AI could also target much more powerful computing resources, including advanced processors in data centres.
The risk reinforces a warning issued last month by Anthropic chief executive Dario Amodei, who suggested that within a year, or even sooner, AI could hijack “the entire internet” using a “persistent botnet”. In a worst-case scenario, the internet could be brought down and paralysed by a malicious botnet.
Experts disagree
The threat has prompted debate among cybersecurity experts, who hold differing views on how serious it is.
Ciaran Martin, former head of the UK’s National Cyber Security Centre, said Anthropic’s report appeared to contain an element of excessive alarm.
“‘We managed to run a small model on a cheap television’ does not mean a global security crisis has occurred,” Martin said.
By contrast, Zhengxiong Li, an assistant professor of computer science and engineering at the University of Colorado Denver, said the prospect of AI becoming an autonomous programme capable of self-replication and independently hijacking devices remained worrying.
“If the programme runs on many devices without relying on a central service, there is probably no single point of leverage to stop it,” Li said.
Jessica Ji, a senior research analyst at Georgetown University’s Center for Security and Emerging Technology (CSET), said the possibility of AI spreading autonomously remained worthy of attention, particularly following a series of disclosures about malicious AI activity over the past three months.
However, she stressed that the likelihood of such a threat could not yet be established with certainty. “We do not yet know exactly how likely this is to happen. The possibility is not zero, but nor is it guaranteed,” she said.
Source: jawawa.id (https://jawawa.id/newsitem/internet-at-risk-of-total-collapse-it-is-only-a-matter-of-time-1791978635)
Originally published by CNBC Indonesia (https://www.cnbcindonesia.com/tech/20261014143346-37-776127/internet-terancam-lumpuh-total-tinggal-menunggu-waktu), translated from Indonesian