Google has unveiled a new experimental artificial intelligence model designed to enhance cybersecurity workflows
specifically focusing on incident response and threat analysis. Developed by the tech giant’s Mandiant threat intelligence division, the AI model—named Sec-Gemini v1—is part of Google’s growing effort to integrate advanced AI into security operations.
Sec-Gemini v1 blends the power of Google's Gemini large language model with near real-time security data and a range of internal tools. These include Google Threat Intelligence (GTI), the Open-Source Vulnerability (OSV) database, and other proprietary resources. Google says this synergy allows the model to excel in areas such as incident root cause analysis, threat attribution, and understanding the impact of vulnerabilities.
The company claims that Sec-Gemini v1 surpasses competing models in multiple industry-standard benchmarks. Notably, it leads by 11% on the CTI-MCQ threat intelligence benchmark and by 10.5% on the CTI-Root Cause Mapping benchmark, which measures how well an AI system can interpret and categorize vulnerability descriptions using the Common Weakness Enumeration (CWE) taxonomy.
In one example provided by Google’s security team, Sec-Gemini v1 successfully identified Salt Typhoon as a known threat actor and presented detailed background, including related vulnerabilities and associated risks. This level of insight is largely credited to its direct integration with Mandiant’s extensive threat intelligence data.
To encourage further development and feedback, Google announced that Sec-Gemini v1 will be made freely available to selected researchers, cybersecurity professionals, academic institutions, and non-governmental organizations (NGOs) for testing purposes.
Found this article interesting? Follow us on X(Twitter) and FaceBook to read more exclusive content we post.
