When AI Shuts Down a Wind Farm for $8
In a security experiment, AI agents managed to shut down a simulated wind farm in less than two hours. The API costs for one of the attacks were just $8. The experiment illustrates how AI can lower the cost and barriers to cyberattacks on critical infrastructure.
In a security experiment, Shinobi Security investigated how independently AI agents could attack a wind farm. The researchers used a simulated facility rather than operating infrastructure. [1][2]
Two AI models achieved the objective of disconnecting the simulated wind farm from the power grid. Claude took 66 minutes, with API costs of approximately $20. Kimi took just under two hours, with API costs of approximately $8. [1][2]
The agents exploited existing credentials, misconfigurations and default passwords, among other weaknesses. The experiment therefore does not necessarily reveal new vulnerabilities. Instead, it demonstrates how AI can automate and combine known attack techniques. [1][2]
What stands out is not only the speed, but also the price: When complex attack sequences can be automated at low API costs, an important barrier to cyberattacks on critical infrastructure is reduced.
The experiment took place under controlled conditions. It does not prove that every real wind farm could be shut down. It does, however, show how AI can exploit existing weaknesses automatically and at low cost.
Another indication of a post-generative present in which AI systems are becoming increasingly autonomous, and our critical infrastructure must be prepared for this new reality.
Sources
This indication by Benedikt Brünner is licensed under
CC BY 4.0
Comments 0
Comments are moderated. New comments appear after review by the editors.
No comments yet.