SentinelOne’s new benchmark, built on the Fast16 case, shows which AI models can sustain a malware investigation and which cannot.
The post Nuclear-Sabotage Malware Benchmark Trips Up Most Frontier AI Models appeared first on SecurityWeek.