AI Models Test Vulnerable to Cyberattacks
Researchers found that several AI models are capable of creating actionable cyberattack plans against industrial systems.
Updated on Oct. 21, 2026 in Artificial Intelligence
Live Poll
Do you trust that current security measures can protect our power and water systems from AI-driven threats?
Frenos researchers conducted nearly 17,000 tests on 10 AI models to evaluate their ability to plan cyberattacks on industrial infrastructure. The study revealed that three models successfully generated attack plans in 90% of their attempts.
Why it matters
The research highlights security concerns as industrial operators face significant budget and staffing constraints when attempting to update vulnerable systems. Experts warn that increasingly capable AI models could soon make such digital attacks easier to execute.
Frenos researchers performed 17,000 tests on 10 AI models, including 2,520 attempts to construct full attack plans. Success rates varied significantly, with three models succeeding 90% of the time, while frontier models ranged from 17.9% to 31.9% success.
The players
Frenos
Frenos is a research organization that evaluates the security risks and capabilities of advanced artificial intelligence models.
The details
The testing process involved providing AI models with data regarding potential targets and tasking them with identifying specific steps to achieve operational objectives. This streamlined setup allowed researchers to measure the models' capacity to orchestrate complex cyber threats against critical systems.
Timeline
October 2026: Frenos released the research findings.
The Tech Race
This study demonstrates how large language models are evolving from simple text generators into tools capable of complex cyber exploitation. It marks a significant shift as researchers now prioritize testing how these models interact with critical infrastructure legacy systems.
The findings suggest that organizations managing essential services may need to accelerate security upgrades to defend against AI-assisted threats. For users, this emphasizes the growing importance of securing interconnected industrial and utility systems against automated exploitation.
The takeaway
The study serves as a critical reminder that AI capabilities are outpacing current industrial security protocols. System operators should prioritize updating defense strategies to mitigate risks posed by automated, model-driven cyber threats.
Further reading
Learn more about the latest developments in Artificial Intelligence.
Live Poll
Do you trust that current security measures can protect our power and water systems from AI-driven threats?
