Rubrik's Lessons From One Month With Mythos Preview
TechTarget, Tuesday, August 19th, 2025
AI found vulnerability chains humans missed, then buried the security team in findings it could not triage.
After joining Anthropic's Project Glasswing in June, Rubrik found Mythos Preview surfaced complex vulnerability chains that conventional scanning and human review missed, including cross-component relationships no single engineer would catch
The volume then became the bottleneck: the existing team could not review and prioritize findings at the pace the AI produced them. The instinct was to hire more reviewers, but CTO and co-founder Arvind Nithrakashyap concluded human-driven remediation could not keep up with AI-speed discovery.
Rubrik instead built a cross-functional team to create automated harnesses managing tool calls and checkpoints. The lesson is that the discovery-to-remediation pipeline itself has to become automated rather than human-gated.