Google Launches A Cheaper Alternative To Large Ai Security Models Like Mythos

5 jam yang lalu

ARTICLE AD BOX

Emma Roth

is a news writer who covers nan streaming wars, user tech, crypto, societal media, and overmuch more. Previously, she was a writer and editor astatine MUO.

Google is launching an AI information exemplary dedicated to quickly uncovering and patching information vulnerabilities. In a blog station connected Tuesday, Google describes Gemini 3.5 Flash Cyber arsenic a “cost-efficient and highly tin alternative” to larger, much costly AI systems, specified arsenic nan 1 offered by Anthropic’s Mythos.

The caller exemplary is built upon Gemini 3.5 Flash and will beryllium disposable first to governments and trusted partners via CodeMender, Google’s security-focused coding agent. As noted by Google, CodeMender tin telephone upon 3.5 Flash Cyber “multiple times astatine precocious velocity and debased cost,” allowing nan AI agents to scan much codification paths and find vulnerabilities.

Anthropic’s Mythos 5 is simply a powerful AI information exemplary released arsenic portion of nan company’s Project Glasswing initiative. The compute-heavy exemplary is costly to use, costing doubly arsenic overmuch arsenic Claude Opus 4.8. Microsoft, which adopted Mythos for its information checks, had its biggest Patch Tuesday this month aft using AI to find vulnerabilities. Other companies, for illustration Google, are racing to support up, while China’s Z.ai claims its exemplary can compete pinch Mythos.

Google says 3.5 Flash Cyber achieved “competitive performance” compared to “significantly larger models” connected nan CyberGym AI cybersecurity benchmark erstwhile called upon up to 5 times. It besides identified 55 “unique confirmed issues” successful nan V8 JavaScript Engine, compared to 47 recovered by Gemini 3.5 Flash and 36 by Opus 4.6. Google adds that 3.5 Flash Cyber recovered 10 issues that nary different exemplary discovered, noting that nan exemplary continued to find caller codification paths and vulnerabilities aft being invoked aggregate times.

Follow topics and authors from this communicative to spot much for illustration this successful your personalized homepage provender and to person email updates.