Jakarta, CNBC Indonesia \u2013 Hackers needed only a brief instruction to\ndeploy artificial intelligence (AI) agents and breach the websites of\nmore than 100 companies. The attacks were carried out almost entirely\nautomatically, with little human intervention.<\/p>\n
The Mandarin-speaking hacker used AI agents to search for security\nvulnerabilities and infiltrate company websites over several hours.\nAmong the victims were agricultural equipment manufacturer Deere,\nhousehold appliance company SharkNinja and engine manufacturer\nCummins.<\/p>\n
Snorre Fagerland, a principal threat researcher at cybersecurity firm\nNorForge, uncovered the activity after finding an open-source AI\nplatform called Cairn being used for hacking.<\/p>\n
In his report, Fagerland said the hacker began the operation on 25\nSeptember. The hacker gave the agents brief instructions in Mandarin,\nfewer than 100 characters long, to direct them to carry out the\nattacks.<\/p>\n
By last week, the agents had attempted to breach more than 1,200\nwebsites in the United States, Canada, Australia and Japan. Successful\nintrusions generally took about 23 hours to identify and exploit a\nsecurity vulnerability.<\/p>\n
Fagerland estimated that the operation cost about US$5,000, or\napproximately Rp90 million, in AI tokens \u2013 around US$50 (Rp902,000) for\neach successful breach. The system is believed to have used a\ncombination of China\u2019s DeepSeek technology and Anthropic\u2019s Claude\nCode.<\/p>\n
In one case, an AI agent failed to find a weakness in a major\ncompany\u2019s public-facing application. It then searched digital records\nfor related infrastructure, identified an internet-accessible control\npanel, and exploited a flaw in data processing that allowed the server\nto run malicious commands.<\/p>\n
\u201cThis is not something for which it is easy to find a solution,\u201d\nFagerland said, as quoted by Reuters on Thursday (22 October 2026).\n\u201cThis is not the work of amateur hackers,\u201d he added.<\/p>\n
He said he had previously expected AI agents used for hacking to\nrequire human supervision. But the findings showed the technology had\nadvanced beyond his expectations.<\/p>\n
\u201cIt\u2019s a new world now,\u201d Fagerland said.<\/p>\n
Deere said it had taken the necessary remedial steps and found no\nindication that sensitive information had been accessed. SharkNinja said\nthe attack targeted a separate website containing only product support\ninformation, and that the threat had been contained.<\/p>\n
Cummins said it was investigating reports of a suspected cyberattack\non a website operated by one of its third-party vendors.<\/p>\n
Reuters could not independently establish the extent to which the\nattacks resulted in unauthorised data extraction.<\/p>\n
The findings add to concerns about the capabilities of autonomous AI\nagents, which can identify vulnerabilities and carry out intrusions,\npotentially stealing data without requiring human direction at every\nstage.<\/p>\n
\u201cDevelopments in this field are truly astonishing,\u201d Fagerland\nconcluded.<\/p>", "url": "https:\/\/jawawa.id\/newsitem\/one-click-by-hackers-rp90-million-cost-1-200-websites-breached-1792651896", "image": "" }, "sponsor": "Okusi Associates", "sponsor_url": "https:\/\/okusiassociates.com" }