PCMag's reporting frames the CMID as a device-level identifier that survives sign-outs, new Microsoft accounts, and even Windows reinstalls without a disk wipe. The article emphasizes that the federal indictment confirmed Microsoft used the CMID to link a suspect's Outlook account to a separate Microsoft account he had used months earlier, undermining the common assumption that fresh accounts are uncorrelated.
By submitting the story and driving it to 156 points, ifh-hn signaled that the technical community sees the CMID revelation as a meaningful privacy story. The high engagement suggests developers view this as a substantive change to Windows' identity model, not just a rediscovery of a known telemetry field.
The editorial argues that malware analysts, pentesters, journalists, and researchers who rely on fresh Microsoft accounts on fresh VMs to stay uncorrelated with their primary identity are operating on a false premise. Because the CMID persists across account changes and non-wipe reinstalls, any workflow assuming a burner account is truly isolated now needs to be revisited.
The editorial notes that Microsoft's public documentation describes CMID-adjacent identifiers with anodyne language about reliability and licensing, but the court filing demonstrates the ID is actually usable to join activity across different Microsoft accounts on the same machine. The refusal to comment on the record, paired with historically broad interpretations of that language, suggests the disclosure gap is deliberate rather than accidental.
A PCMag story picked up by Hacker News (156 points) surfaces a detail buried in a US federal indictment: Microsoft has been able to tie individual Windows machines to real people using a persistent identifier called the Client Machine ID (CMID). The document lays it out plainly — investigators asked Microsoft for records tied to a suspect's Outlook account, and Microsoft returned not just IP logs but a device-level ID that also matched activity on a *different* Microsoft account the same suspect had used months earlier.
The CMID is generated at Windows setup, stored locally, and transmitted to Microsoft when the machine talks to first-party services like OneDrive, Store, Update, or a signed-in Microsoft account. It is not the hardware UUID from the motherboard, not the TPM endorsement key, and — critically — not something that resets when you sign out, create a new Microsoft account, or reinstall Windows without wiping the disk. It behaves like a browser supercookie, but for the operating system.
Microsoft did not comment on the record for the PCMag piece. The company's public documentation describes CMID-adjacent identifiers as being used for "service reliability" and "licensing," language that has historically covered a wide range of cross-service joins. What's new here isn't the existence of a device ID — every major OS has one — it's the confirmation, in a court filing, that Microsoft will hand it over and that it survives the reset ceremonies most users assume start them from a clean slate.
If you build software on Windows, or ship endpoints running it, this changes the shape of a few threat models you probably haven't revisited in a while.
The first is the "burner account" fiction. A lot of security-adjacent work — malware analysis, competitor research, pentesting, moderation, journalism — relies on the idea that a fresh Microsoft account on a fresh VM is uncorrelated with your primary identity. The indictment shows that Microsoft can and will bridge those accounts server-side using the CMID. The reset button in Windows Settings does not reset the CMID; only a full disk wipe plus a fresh OS install with a new hardware fingerprint does. VM snapshots don't help either — cloning a VM clones the CMID, which is arguably worse: now two "different" identities share one device fingerprint that Microsoft sees as identical.
The second is corporate telemetry hygiene. Enterprises spend real money on Group Policy configurations that disable diagnostic data, block the Microsoft Store, and gate OneDrive. Most of those policies target the *diagnostic* channel — the one Microsoft publishes schemas for. The CMID appears to ride on service-authentication traffic, which those policies generally don't touch, because breaking it would break Windows Update. You can turn off telemetry and still be identifiable.
The third is the regulatory picture in the EU. GDPR treats persistent device identifiers as personal data when they can be linked to a natural person, and the Digital Markets Act imposes additional constraints on gatekeepers combining data across services without explicit consent. A CMID that lets Microsoft join an Outlook account to a Store account to a Windows install is exactly the pattern regulators wrote those rules against. The precedent from the CNIL's Google Analytics rulings and the Meta cross-service consent decisions suggests this is not academic. Expect a European DPA to open a file on this within the quarter.
The fourth, and the one that matters most for practitioners: you probably have Microsoft accounts glued to infra you'd rather not have glued to your name. GitHub is a Microsoft property. So is npm. So is VS Code's sync. If a CMID join can travel from Outlook to a Store install, there's no architectural reason it can't reach any other Microsoft-authenticated surface — and the indictment is a proof of existence that the plumbing is in place and the legal team is willing to use it.
For most senior developers, the practical response is smaller than the headline suggests, but it isn't zero.
If you run analysis VMs, treat the CMID like you already treat cookies: assume it's persistent, assume it's linkable, and rebuild the machine from a clean ISO — not a snapshot — between engagements. Cloning a golden image and letting each analyst sign into their own MS account inside the clone is now a correlation vector, not an isolation strategy. On Hyper-V and VMware, generate new hardware UUIDs on each provision, and skip the "link this device to your Microsoft account" step during OOBE. Local accounts are still creatable via `Shift+F10` → `oobe\BypassNRO` at setup, though Microsoft has been narrowing that path with each Windows 11 release.
For CI runners on Windows — GitHub Actions `windows-latest`, Azure DevOps hosted agents, self-hosted Windows fleets — the risk surface is different. Hosted runners are ephemeral and don't sign into Microsoft accounts, so the CMID doesn't have a personal identity to bind to. Self-hosted runners that log in with a service account inherit that account's CMID history. If you've been reusing the same Windows AMI for two years, every job that ran on it shares one device fingerprint on Microsoft's side. That's fine until it isn't.
For product teams shipping Windows desktop software: this is a nudge to audit what you send back to Microsoft's advertising ID and app-usage APIs. The CMID story will make privacy-conscious users more suspicious of *any* Windows-side identifier, and "we only use it for crash reporting" is going to land worse than it did last week.
The useful frame here isn't outrage — device IDs exist on macOS, iOS, Android, and ChromeOS too — it's that the reset ceremonies we've been teaching each other for a decade quietly stopped working, and nobody sent a memo. The takeaway isn't "stop using Windows"; it's that "new account on the same machine" is now a documented correlation, not a privacy boundary, and your runbooks should say so. Expect Microsoft to publish a clarifying blog post within two weeks that reframes the CMID as a benign licensing artifact, expect at least one EU regulator to disagree, and expect the next major Windows 11 feature update to quietly add a way to rotate it — which will itself become the next story.
Well they can’t use that to track users of Linux.I was a big fan of Microsoft ten to fifteen years ago. I’ve since transitioned my whole family off Microsoft products now over to Linux, Apple, and proton. Edit: and Brave.I really thought their corporate culture would’ve changed after the late 90’s b
To me this indicates that Microsoft has some sort of traffic analysis performed on endpoints, then linked to GDID. I'd guess this is part of Defender's real time protection or MAPS.Fun fact, Microsoft Defender MAPS was previously named SpyNet.https://en.wikipedia.org/wiki&#x
I guess we’ll see a Windows tool that sets your identifier to this suspect’s “g:6755467234350028” very soon (weird ID, by the way. 16-digits makes sense, but I would have expected it to be hexadecimal)Also, can anybody tell how “Microsoft had records showing that on May 12, 2025, at 19:21 UTC, the G
The criminal complaint missing in this article: https://www.justice.gov/usao-ndil/media/1450651/dlBasically, Microsoft logs things from windows users, including and not limiting to, the machine GDID, IPs that come with the GDID, and when and what exact URLs accessed. So
Top 10 dev stories every morning at 8am UTC. AI-curated. Retro terminal HTML email.
The interesting part is not really the existence of a machine identifier. Almost every modern OS has some equivalent. The bigger question is the boundary: which components can access it, and when does a local identifier become a remote tracking identifier? A machine-id sitting on disk is very differ