Google Unveils Gemini 3.5 Flash Cyber to Strengthen AI-Powered Cybersecurity and Software Protection

Gemini 3.5 Flash Cyber, a new AI model from Google, is intended to enhance cybersecurity by automatically identifying and resolving software vulnerabilities. According to the company, the new model makes it simpler for businesses to safeguard their software and digital systems because it is quicker and less expensive than massive AI security models.

Gemini 3.5 Flash Cyber has a lightweight design, in contrast with larger AI models that need more processing power and are more expensive. Because of this, it can quickly do several security scans on large software projects without disrupting development. According to Google, this method lowers the possibility of cyberattacks and assists developers in identifying security issues early.

The new AI model integrates with Google's internal security tool, CodeMender. Instead of just checking software once, CodeMender runs Gemini 3.5 Flash Cyber multiple times over different sections of the code. It then compiles all of the findings into a thorough security report, allowing developers to swiftly identify and address critical vulnerabilities.

According to Google, the model performed well on the CyberGym benchmark, which assesses how AI systems discover real-world software security issues. The company stated that Gemini 3.5 Flash Cyber produced outcomes comparable to much larger and more expensive AI models while using fewer resources.

Google also stated that the new AI cybersecurity strategy is already in use across its own services, such as Chrome, Android, Google Cloud, YouTube, and Google Ads. During internal testing, the model apparently discovered major security flaws in public APIs and identified a memory-related vulnerability in a production system in under two hours. It also developed a workable proof-of-concept to demonstrate the security issue, allowing engineers to address it rapidly.

Early testing by cybersecurity company Wiz and Google's Cloud security team showed that Gemini 3.5 Flash Cyber outperforms the normal Gemini 3.5 Flash model in security-related tasks. Its speed and low operating costs make it ideal for daily software upgrades, code reviews, and large-scale development initiatives.

To avoid the potential of misuse, Google is not releasing the model to the public immediately. Gemini 3.5 Flash Cyber will first be accessible in a limited-access pilot program for government organizations and trusted enterprise customers who use CodeMender.

With the release of Gemini 3.5 Flash Cyber, Google expands its AI-powered cybersecurity capabilities. The company believes that the new model would help developers uncover software vulnerabilities faster, increase software security, and strengthen defense against emerging cyber threats while keeping expenses under control.

This article is based on information from Cyber Security News