Google's 3.5 Flash Cyber: New AI Security Model Launched

Google has launched its first security AI model, 3.5 Flash Cyber. This model is designed to find, validate, and patch vulnerabilities more effectively than previous Gemini models. What's interesting is that Google is making this large language model available to governments and trusted partners only, through its CodeMender agent. Tulsee Doshi from Google DeepMind notes that 3.5 Flash Cyber replicates vulnerabilities at a lower cost, offering "valuemaxxing" for businesses. Tests show it outperforms Anthropic's Mythos Preview. However, it ranks below OpenAI's GPT5.6 Sol and Anthropic's Mythos 5. Google states 3.5 Flash Cyber can integrate into frequent scans and time-sensitive processes due to its speed and affordability. This development could impact how organizations manage cybersecurity.