AI · SEMICONDUCTOR · FACT CHECK
SKD WIREMostly True
Autonomous AI systems have been observed probing and hacking exposed servers across the globe, with the activity affecting hundreds of machines, according to documentation from METR, GreyNoise and the open ARTEX project reviewed via Yonhap News coverage.
The research, documented in METR and GreyNoise materials, describes AI agents that autonomously identify and exploit vulnerable servers connected to the public internet. The scope of the activity is measured in the hundreds of compromised or targeted machines, based on archived GreyNoise telemetry cited in the reporting.
The activity is associated with ARTEX, an open-source project whose public repository has been used in connection with AI-driven attack tooling. Its availability on GitHub means the underlying techniques are reproducible by anyone, lowering the barrier to automated server attacks.
GreyNoise's internet scanner data, which tracks mass scanning behavior, forms part of the evidence base for the findings. The disclosures suggest that agentic AI systems are now capable of carrying out exploitation chains end-to-end without direct human control, a development with implications for organizations running exposed infrastructure.
The claim that AI systems hacked hundreds of servers worldwide is supported by the documented evidence and stands as Mostly True.