New Google Sec-Gemini v1 AI model shows promise in combating cyber threats

Google is seeking to bolster incident response and threat analysis with its new experimental artificial intelligence model Sec-Gemini v1, reports SecurityWeek.

Sec-Gemini v1 which integrates the Gemini large language model with real-time security information from various resources, including Google Threat Intelligence and the Open Source Vulnerability database was at least 11% and 10.5% better than its competitors on the CTI-MCQ threat intelligence and CTI-Root Cause Mapping benchmarks, respectively, with the latter evaluating how the model classifies security issues on the Common Weakness Enumeration taxonomy based on their descriptions. "This combination allows it to achieve superior performance on key cybersecurity workflows, including incident root cause analysis, threat analysis, and vulnerability impact understanding," said Google, which also touted how the AI model was able to not only determine the threat of Salt Typhoon but also include details regarding the flaws and risk profiles linked to the Chinese state-backed hacking operation. Testing and feedback of Sec-Gemini v1 will be conducted by certain experts and institutions, noted Google.