Anthropic's Project Glasswing utilizes Mythos Preview, an AI model that found vulnerabilities across major operating systems and browsers, including bugs that survived decades of human audits. The model achieved a 72.4% success rate in Firefox JS shell exploit development and can chain four independent bugs into exploit sequences bypassing browser renderer and OS sandboxing, perform local privilege escalation in Linux through race conditions, and build 20-gadget ROP chains targeting FreeBSD's NFS server. Attackers are using custom MCP servers hosting LLMs for autonomous attacks including backdoor creation, infrastructure mapping, vulnerability assessment, and tool execution for domain admin access. The median time from disclosure to weaponized exploit dropped from 771 days in 2018 to single-digit hours by 2024.