Anthropic, a San Francisco-based AI startup, announced that its forthcoming AI model, Claude Mythos, has successfully identified thousands of software vulnerabilities, some dating back 27 years. Due to the potential risks associated with its capabilities, the company has chosen not to release Mythos publicly. Instead, it will collaborate with cybersecurity specialists and engineers from the open-source community to utilize the model defensively. This initiative, named Glasswing, includes partnerships with major companies such as Amazon, Apple, and Microsoft, as well as cybersecurity firms like CrowdStrike and Palo Alto Networks. Anthropic is investing around $100 million in computing resources for this project. The vulnerabilities discovered by Mythos were often subtle and previously undetected, highlighting the advanced coding capabilities of AI. The urgency of this work is underscored by the rapid pace at which vulnerabilities can now be exploited, with the time frame shrinking from months to mere minutes. Despite a federal directive to terminate contracts with Anthropic, discussions with the US government regarding Mythos are ongoing.