Data resilience company Rubrik Inc. has rebuilt its code review pipeline after utilizing Anthropic’s Mythos Preview model to identify a significant number of potential security issues within its own codebase. The company found that the AI model surfaced so many vulnerabilities that it opted to overhaul its review process rather than scale up human review efforts, as reported by Silicon Angle.Rubrik gained access to Anthropic's Project Glasswing, which provides early access to the Mythos Preview model, a tool designed to find software flaws and construct attack chains. Rubrik assembled a dedicated team to develop a "harness" around the AI model, incorporating business and security context to manage its operations. The initial scanning process was broad, covering entire code repositories, with subsequent passes narrowing focus based on identified patterns. While Rubrik reported a substantial reduction in raw findings to prioritized issues, specific numbers were not disclosed.A key takeaway for Rubrik was the realization that not all aspects of vulnerability remediation should be automated, opting for trustworthy automation over maximum automation for critical issues. This approach aligns with Anthropic's broader findings within Project Glasswing, where partners have identified over 10,000 high- or critical-severity vulnerabilities, though patching rates have been a challenge across various industries, including technology, power, water, and healthcare.Silicon Angle
Source: AI/ML
Rubrik rebuilds code review pipeline after AI model finds numerous security flaws
(Adobe Stock)
An In-Depth Guide to AI
Get essential knowledge and practical strategies to use AI to better your security program.
Get daily email updates
SC Media's daily must-read of the most current and pressing daily news
You can skip this ad in 5 seconds