Google is introducing new Gemini “Flash” AI models, including a specialized cybersecurity system aimed at quickly finding and patching software vulnerabilities. Multiple outlets report that Google plans to roll out “Gemini 3.5 Flash Cyber” (and related Flash versions), positioning the security model as a lower-cost alternative to larger, more expensive frontier AI systems used for similar tasks. Google describes the model as cost-efficient while maintaining strong capability for software security work.
According to the sources, Gemini 3.5 Flash Cyber is designed to scan large codebases, identify critical vulnerabilities, and propose or apply patches. One outlet adds that the model supports multiple rapid scans across extensive repositories, which is intended to reduce the need to rely on higher-cost systems for each security review.
On availability and misuse controls, reports agree that access is restricted. Google is using a limited-access pilot and, at least initially, makes the model available to governments and trusted partners. This approach is intended to prevent misuse of the specialized cybersecurity capability while it is evaluated in controlled settings.
