Istanbul:Some of the world's most advanced artificial intelligence models have successfully developed cyberattack plans targeting critical infrastructure, including water and power systems, in approximately 90% of attempts, according to research published recently.
According to Anadolu Agency – English, cybersecurity company Frenos conducted tests on 10 AI models to assess their ability to identify weaknesses, evaluate access requirements, and plan attacks against industrial systems governing essential services. The study involved 16,980 tests, which included 13,884 individual decisions, 2,520 attempts to develop complete attack plans, and 576 simulated runs.
Three leading commercial AI models produced workable attack plans in 89.5% to 92.9% of planning attempts, whereas other models showed significantly poorer performance. The most successful foreign open-weight model achieved a 36.2% success rate, in contrast to 5.2% for the best-performing US open-weight model. Notably, the researchers chose not to disclose the names of the models tested.
Frenos highlighted that the tests were based on provided information and simulated scenarios rather than actual attacks. The AI models sometimes neglected vital steps or assumed access privileges they did not possess. The research did not prove that AI could independently identify targets or execute cyberattacks.
These findings arise amid increasing concerns over cyberattacks on industrial systems, such as a recent case in France where hackers manipulated refrigeration systems at numerous supermarkets, leading to temperature spikes, food spoilage, and reportedly causing fires. Frenos also referenced a January cyberattack on a Mexican water utility where attackers used Anthropic's Claude AI to identify an industrial monitoring interface and suggest further access attempts, though these attempts ultimately failed, and no operational control systems were compromised.
The company is advocating for critical infrastructure operators to enhance network separation, restrict remote access, and vigilantly monitor suspicious activities to mitigate the risk of AI-assisted cyberattacks.
Source Anadolu Agency – English