A security researcher has published details of 11 unpatched vulnerabilities in Tailscale and Headscale, the mesh VPN software thousands of small teams use to connect laptops, servers, and cloud instances without running a traditional VPN gateway. The report, nicknamed "Tailworm" by its author Erica Windisch of Cyberdione Labs, was tested against Tailscale 1.104.2 and the open-source Headscale 0.29.5 and covers five areas: the built-in SSH server, the "serve" and app-connector features, the DERP relay client, and Headscale itself.
The most serious issue is a race condition in Tailscale's SSH server. Tailscale's background service (tailscaled) runs with root privileges, and when an SSH session forwards a Unix socket, the file and permission checks meant to keep that privileged access contained can be raced and bypassed. In practice, that means an attacker who already controls one device authorised on your network — and who has an SSH access rule pointing at another — may be able to reach root on that other server and even enrol a rogue device of their own. This is lateral movement, not an open door from the internet: the attacker needs a foothold first, typically through phishing, a stolen device, a compromised software package, or a vulnerable app already running on an authorised machine.
As of today, Tailscale has not published a security bulletin, a CVE, or a fixed version. The company's VP of engineering, Mike Shaver, confirmed on Reddit that Tailscale had already been investigating the same areas with outside security specialists and is working on fixes, while adding that the company is not telling everyone to disable Tailscale SSH outright. The researcher says she shared the report and proof-of-concept code with Tailscale's security contact before publishing.
What this means for your business
If your team, or a contractor who manages your servers, uses Tailscale or Headscale to connect machines remotely, this is worth 20 minutes this week rather than an emergency. First, review which devices are actually authorised on your network and remove anything stale — an old laptop, a former employee's phone, a test server nobody uses — since the bug needs an existing foothold to work, and a cleaner device list shrinks that attack surface directly. Second, if you use Tailscale SSH, turn on its "check mode" (it re-authenticates SSH sessions rather than trusting network membership alone) rather than switching the feature off entirely. Third, watch for Tailscale's official advisory and patched release and apply it promptly once it ships — treat it like any other pending vendor patch, not a reason to panic in the meantime.
More broadly, this is a useful reminder for any business relying on a mesh VPN or "zero trust" network: being on the network is not the same as being trustworthy, and the easiest way for an attacker to get the first foothold these bugs need is a device nobody remembered to remove.