The cybersecurity landscape is shifting with the emergence of Mythos, an AI developed by Anthropic capable of identifying significant vulnerabilities in software rapidly. This model, which was inadvertently exposed last month, is believed to have uncovered thousands of high-severity flaws that could potentially be exploited by malicious actors.
Initially, Anthropic planned to keep Mythos confidential due to its hacking capabilities, indicating that the implications for public safety and national security could be serious. However, the company has since opted to share the technology with select firms, including Amazon Web Services, Apple, and Google, under a program named Project Glasswing. This initiative aims to enable these organizations to identify and rectify vulnerabilities in their own systems proactively.
Reports have emerged of unauthorized access to Mythos by individuals on a private forum, suggesting potential oversights in security measures. Despite initial intentions to keep the AI under wraps, its testing by cybersecurity experts is garnering considerable interest as they explore its capabilities.