Rubrik Expands Project Hourglass With AI Code Guardian Security Tool

Cybersecurity company Rubrik is expanding its AI-powered security initiative, Project Hourglass, with Code Guardian, a tool designed to identify vulnerabilities in software and help enterprises prepare for advanced AI-driven cyberattacks. The October 9 announcement also brings additional technology partners into the initiative and expands access to Rubrik Agent Cloud.

Code Guardian uses Anthropic’s Claude Mythos 5 through a custom security framework to analyse code repositories in isolated environments. The approach is designed to help security teams identify realistic attack paths, validate vulnerabilities and prioritise remediation before attackers can exploit them.

Scanning Code in Isolated Environments

Unlike conventional code-scanning tools that may generate large numbers of alerts, Code Guardian focuses on identifying connected vulnerabilities that could form a practical attack path. It analyses relationships across files, services, authentication mechanisms and cloud boundaries to determine how weaknesses might be exploited together.

The system works on secure, isolated copies of source-code repositories rather than directly examining live production systems. This separation is intended to reduce operational risk while allowing security teams to conduct more advanced testing.

Anthropic’s AI Supports Vulnerability Detection

Code Guardian integrates Claude Mythos 5, an advanced AI model used to identify software vulnerabilities and assess potential attack chains. Rubrik's security framework adds verification steps to distinguish exploitable issues from findings that may not represent a practical threat.

Once critical vulnerabilities are confirmed, the tool can provide remediation guidance and route findings into developer workflows through platforms such as Jira and GitHub. This is intended to help organisations move from detecting security problems to addressing them more efficiently.

Project Hourglass Expands Its Partner Network

As part of the October 9 expansion, Rubrik added AHEAD, Trace3 and WWT to Project Hourglass. They join existing participants including Cognizant, Deloitte, LTM, HCLTech, NTT DATA and Wipro.

The initiative is designed to help enterprises strengthen cyber resilience as AI tools become more capable of discovering and exploiting software weaknesses. The expanded partner network will support efforts to test code, identify risks and improve security practices across enterprise environments.

Why AI Code Security Matters

AI-generated code can accelerate software development, but it can also introduce vulnerabilities that are difficult to identify through conventional testing alone. As organisations adopt AI coding assistants and autonomous agents, security teams face pressure to assess larger volumes of code and investigate increasingly complex attack paths.

Tools such as Code Guardian aim to help defenders keep pace by combining AI-assisted analysis with vulnerability validation and recovery processes. Rubrik's approach reflects a broader cybersecurity shift towards identifying realistic attack scenarios and prioritising weaknesses according to their potential impact.