Rubrik Adapts Workflows to AI-Driven Vulnerability Discovery with Mythos Preview

AI Exposes Capacity Gaps

After gaining access to Anthropic's Mythos Preview through Project Glasswing, Rubrik discovered that the AI model could identify complex vulnerability chains that its conventional security tools and methodologies missed. These findings included relationships between components across large codebases that were not detectable by standard scans or individual engineer reviews. This surge in identified issues created a prioritization bottleneck for Rubrik's engineering team.

Shifting from Human to Automated Remediation

Rubrik initially considered hiring more human reviewers to handle the increased volume of potential issues. However, the company quickly abandoned this plan, realizing that human-driven remediation could not keep pace with the speed of AI-driven discovery. This led to a strategic pivot towards automation.

Developing an Effective Harness for AI Findings

Rubrik assembled a multi-functional engineering and infosec team to build an automated system for high-fidelity threat discovery and elimination. The focus was on creating a software layer around Mythos to manage tool calls, checkpoints, and add business and security context. This system aims to reduce the number of findings that ultimately require human engineer review and remediation.

Refining the Discovery Process

Rubrik's new workflow involves using Mythos for an initial whole-repository scan. The insights from this scan then inform subsequent, more targeted passes. These progressive passes are designed to filter out noise, ensuring that only high-quality findings are routed to the appropriate teams and prioritized effectively for action.