Claude and Kimi take down a simulated wind farm in test

On 2026-11-13 at 19:30, Golem.de reported that in a test environment, the AI models Claude and Kimi succeeded in shutting down a simulated wind farm in a creative way. The report tied the finding to KI and to a security gap. It did not disclose the specific attack path used, nor the conditions under which the test was run, leaving open how the evaluation was set up, who took part, and what criteria were applied to the outcome. Based on the single report available so far, the episode is still limited to the public disclosure of this test result. No further verifiable details have emerged about the exact methods the two models employed or the surrounding test parameters, and the accounts provided do not elaborate on whether the shutdown was intentional, incidental, or part of a broader assessment. For now, the timeline of the incident rests solely on this one publication, with the underlying technical steps and testing setup remaining unspecified in the available material.