Anthropic's AI Model Targets Cybersecurity Vulnerabilities

Anthropic’s latest push into the security space arrives under a veil of secrecy. The company announced a new AI model—dubbed Mythos Preview—intended to hunt for software flaws, yet the public rollout comes with few concrete performance metrics. In its brief blog note, Anthropic claimed the system has already identified “thousands of high‑severity vulnerabilities” across a range of recent codebases, suggesting a level of efficacy that could attract enterprise buyers.

However, when pressed for specifics, the firm’s own cyber lead chose not to elaborate. This reticence raises questions about how much of the model’s success can be independently verified and what the practical implications are for organizations looking to bolster their defenses. The following exchange with The Verge captures that tension, as Newton Cheng, head of Anthropic’s frontier red team, walks a thin line between acknowledging early results and withholding the granular details that would let analysts assess the tool’s true impact.

*In an interview with The Verge, Newton Cheng, the cyber lead for Anthropic's frontier red team, declined to share specific details of the model's cybersecurity successes, but Anthropic's blog post said that in recent weeks, Mythos Preview has flagged "thousands of high‑severity vulnerabilities, incl*