Google announces new AI model for detecting security vulnerabilities
Gemini 3.5 Flash Cyber offers a cost-effective alternative to large, expensive security models.
1Google's new security-focused AI model
Google unveiled Gemini 3.5 Flash Cyber, a new AI security model designed as a cost-efficient alternative to larger, more expensive systems. The model specializes in rapidly identifying code vulnerabilities and security flaws in software applications.
2Integration with CodeMender
The new model will work with Google's security-focused coding agent CodeMender. Optimized for high-speed, low-cost multiple invocations, the system can scan more code paths to uncover potential security threats efficiently.
3Competitive benchmark performance
Google claims its model achieved competitive results against 'significantly larger models' in CyberGym benchmark tests. It identified 55 unique issues in V8 JavaScript Engine, outperforming rival solutions from competitors.
4Challenging Anthropic's Mythos
Positioned as a competitor to Anthropic's expensive Mythos model, Google's solution initially targets governments and trusted partners. Microsoft's recent massive Patch Tuesday highlights the growing importance of AI-powered security solutions.
5Budget-friendly options
Google also announced 3.5 Flash-Lite as its 'most cost-effective' model and upgraded versions of 3.6 Flash. The company aims to capture market share by offering solutions for different budgets and security needs.
6AI arms race in cybersecurity
Google's move underscores the increasing importance of AI in cybersecurity. With competitors like China's Z.ai developing similar models, the industry is poised for intensified competition in AI-powered security solutions.