deepmind.googlearticleIntroducing Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber
Google DeepMind's announcement of Gemini 3.6 Flash and 3.5 Flash-Lite also introduces Gemini 3.5 Flash Cyber, a smaller model fine-tuned for finding and fixing code vulnerabilities that runs as multiple coordinated agents inside CodeMender, Google's automated code-security agent. The post reports competitive-with-frontier results on the CyberGym benchmark and states the model will be released only to governments and trusted partners via a limited pilot, citing dual-use risk since AI is now finding vulnerabilities faster than defenders can patch them.
TL;DR DeepMind pairs a new specialized model (3.5 Flash Cyber) with its CodeMender agent for automated vuln finding/patching, gated to a government/trusted-partner pilot.
Also seen from another angle🧠 AI →
Technologies named here
Techniques named here
- Finding and fixing cybersecurity vulnerabilities at scale
Related in AI Security
blog.talosintelligence.comtoolIntroducing CAIRN: Frontier tracking for AI-integrated malware (Cisco Talos)
github.comadvisoryFlowise MCP Security Bypass Enables RCE
github.comrepoYesWeHack Claude Kit
jfrog.comarticle