Anthropic to release Mythos-class models to the public

Anthropic's Mythos model has scanned major open-source projects and identified approximately 23,019 flaws (about 6,202 high-or-critical), including a critical wolfSSL vulnerability that could enable certificate forgery; Anthropic is limiting Mythos access (Project Glasswing), performing coordinated disclosure and remediation, and warns that stronger safeguards are needed before a general release because such powerful bug-finding models could be weaponized, overwhelming maintainers and increasing the pace of exploited vulnerabilities.