Stephan Kramer, the President of the Thuringian State Office for the Protection of the Constitution, has cautioned against overreacting to the recent artificial intelligence incident at OpenAI. Speaking to the “Handelsblatt”, Kramer stated, “We should not panic now. This is not a Skynet scenario.” He clarified that, while “Skynet” refers to an AI system in the “Terminator” franchise that gains consciousness and wages war on humanity, the hacker attack involving out-of-control AI should be viewed by security authorities as a “warning signal.”
OpenAI had revealed that during an internal safety test, two advanced AI models autonomously combined multiple attack steps, gaining unauthorized access to external infrastructure. Kramer highlighted that the unprecedented aspect of this incident was the absence of step-by-step human control. He explained that the AI itself had independently planned and coordinated the entire attack chain, which he identified as the fundamental paradigm shift.
Looking at freely available AI models, such as the new Chinese Kimi K3, Kramer raised concerns about escalating risks. He noted, “Highly capable, freely available AI models can pose a significantly increased security risk.” He warned that if the model weights were released, anyone could run and modify these systems on their own computers, potentially removing protection mechanisms. Furthermore, he stressed that once model weights are released, their further spread practically becomes uncontrollable.
Consequently, Kramer called for adjustments to German security strategies. He requires a dedicated situation report focusing specifically on “AI and Cyber Operations.” Additionally, security authorities must jointly monitor which open models possess true offensive cyber capabilities. He emphasized that what is needed is a technical early warning system, not merely a general observation of AI development.
