Data security company Rubrik has secured early access to Anthropic's Mythos Preview model through the Project Glasswing program. The AI model specializes in identifying software vulnerabilities and mapping potential attack chains that adversaries might use to compromise systems.
Anthropic developed Mythos Preview as part of its efforts to apply artificial intelligence to cybersecurity challenges. The model analyzes code and system configurations to detect security weaknesses that traditional scanning tools might miss. Project Glasswing serves as Anthropic's early access program, allowing select partners to test and provide feedback on the technology before wider release.
The Mythos Preview model works by examining software for common vulnerability patterns and then simulating how attackers could chain multiple flaws together to achieve deeper system access. This approach mirrors real-world attack methodologies, where adversaries rarely rely on single vulnerabilities but instead combine multiple weaknesses to bypass security controls. The AI can process large codebases and complex system architectures faster than manual security reviews.
For Rubrik customers, this integration could mean more comprehensive security assessments of their backup and data recovery infrastructure. The ability to identify attack chains before they are exploited gives defenders a significant advantage in hardening systems. However, the effectiveness of AI-driven vulnerability detection depends on the model's training data and its ability to adapt to new attack techniques.
Security teams should view AI-assisted vulnerability scanning as a complement to, not a replacement for, existing security practices. Organizations should continue maintaining robust patch management programs, conducting regular security audits, and implementing defense-in-depth strategies. While AI models like Mythos Preview can accelerate flaw discovery, human expertise remains essential for prioritizing remediation efforts and understanding the business context of security risks.
Source: https://www.scworld.com/brief/rubrik-rebuilds-code-review-pipeline-after-ai-model-finds-numerous-security-flaws