Meta Rushed to Fix Muse ‘VM Escape’ Vulnerability Soon Before Launch

In the weeks before the launch of Meta’s Muse AI agent, engineers discovered several security vulnerabilities — including at least one that could have let users break out of their per-user virtual machines and reach Meta’s own sensitive internal databases, 404 Media reports. The flaws were severe enough to reach Mark Zuckerberg, triggering a multi-team ‘mad dash’ to fix ‘a sudden spike in reported KVM escapes’ before release. One Meta source described ‘half-baked protections being rushed out to enable the launch,’ with senior engineers warning a major breach was inevitable.