Your Encryption Keys Are in Virginia: On BitLocker, the FBI, and Why European Universities Need Sovereign Software
Abstract
Forbes reported last week that Microsoft gave the FBI BitLocker recovery keys; Microsoft told Forbes it sometimes provides such keys to authorities. On eligible devices, Windows can back a recovery key up to a Microsoft account during setup. For European universities that handle research data, student records, and HR information under GDPR, this is not an abstract concern. It is a structural problem. The answer is not a technical workaround. It is sovereign, publicly funded, openly licensed software — and a principle that the EU has articulated but not consistently practised: public money, public code.
Contents
The Story
Last week Forbes reported that Microsoft gave the FBI the BitLocker recovery keys for three laptops, under a warrant according to local reporting in Guam; Microsoft told Forbes that it sometimes provides such keys to authorities and receives about 20 such requests a year [[1]] [[2]]. The underlying case was a fraud investigation in Guam involving Pandemic Unemployment Assistance [[2]]. The laptops were encrypted with BitLocker — the full-disk encryption built into Windows, which many institutions and individuals rely on as their primary protection against unauthorised data access.
The mechanism depends on configuration. Microsoft’s documentation says a recovery key may be saved to a Microsoft account, a work or school account, a USB drive, a file, or a printout [[3]]. On a device that qualifies for device encryption, Windows encrypts the drive by itself and completes the protection once the recovery key is backed up: to Microsoft Entra ID or Active Directory if the device is joined to one, otherwise to the online Microsoft account of an administrator who signs in with one [[4]]. Microsoft calls the Microsoft account “the default recommended recovery key storage method” for devices joined to neither [[5]], and since Windows 11 version 24H2 more devices qualify [[4]]; managed organisations can instead control escrow through policy. In the reported Guam case Microsoft held the relevant keys and produced them; where a user never uploaded the key, there is nothing to produce. That does not mean every modern Windows installation uploads a key to Microsoft.
Matthew Green, the Johns Hopkins cryptographer, told Forbes what the lesson is: “if you have access to keys, eventually law enforcement is going to come” [[1]]. According to heise, Jennifer Granick of the ACLU pointed out to Forbes that this data is likely of great interest to many governments [[6]]. The same heise report cites Microsoft’s transparency report: 5,296 requests from German law enforcement for July to December 2024, for Microsoft services in general rather than BitLocker keys, and “Not every request automatically results in the disclosure of data” [[6]]. As I read it, the same mechanism is available to any government that can issue a Microsoft-compatible legal order — not only the US Department of Justice.
That last point is the one European institutions should be reading carefully.
Why This Is a European Problem
The CLOUD Act of 2018 amended US law so that a provider served with a qualifying order must disclose data within its “possession, custody, or control, regardless of whether such communication, record, or other information is located within or outside of the United States” (18 U.S.C. § 2713) [[7]]. The order, provider, account, and available legal challenges still matter. Foreign server location alone is not a complete jurisdictional shield. Microsoft’s report for the second half of 2025, later than this post, records that it gave US law enforcement content data of three non-US enterprise customers whose data was stored outside the US, one of them in the EU or EFTA [[8]].
The Guam case shows nothing about location: Guam is US territory, and the reports do not say where the keys were stored. Nor have I found a Microsoft statement of where a Microsoft account’s recovery keys sit. As of October 2026, its privacy statement says that personal data is typically stored “in your region or in the United States, often with a backup to a data center in another region” [[9]]. Under § 2713 that answer would not decide the matter. If a university’s BitLocker recovery keys are held in a Microsoft-controlled account, the institution has created a route by which a US production order may reach them. Fig. 1 sorts the places a key can sit by that test.
This is a conflict-of-laws and risk-management problem, not a finding that every US-cloud use is unlawful. In Schrems II (2020), the Court of Justice invalidated the EU-US Privacy Shield and required case-specific safeguards for certain transfers [[10]]. Since July 2023 a new adequacy decision, the EU-US Data Privacy Framework, has again covered transfers to participating US companies [[11]]. The General Court upheld it against a challenge on 3 September 2025 (Latombe v Commission, T-553/23), and an appeal (C-703/25 P) was lodged on 31 October 2025 [[12]]; I have not checked its status. The framework answers the transfer question, not the provider’s duty under § 2713. Neither Schrems II, the new decision nor the GDPR automatically decides the lawfulness of each institutional deployment. They do make data location, provider control, encryption-key custody, transfer mechanism, and access law questions that the institution must assess together.
European universities hold exactly the kinds of data that make this a real rather than a theoretical concern:
- Research data: medical studies, clinical trials, interviews with human subjects, social science datasets — all subject to strict ethical and legal protections
- Student records: academic performance, personal circumstances, disciplinary proceedings
- HR data: employment contracts, salary records, health information, union activity — particularly sensitive under German and EU labour law
- Correspondence and draft documents: research in progress, grant applications, peer review material
If the disk holding any of this is encrypted with BitLocker and its recovery key is held by a third-party provider, the encryption protects against some threats while leaving a provider-mediated recovery path. Whether a state can obtain the key depends on the applicable legal process and provider control. For organisational accounts, Microsoft says it always tries to redirect law enforcement to the customer and notifies the customer unless the law forbids it [[8]]. Its report for the second half of 2025 counts 190 requests from law enforcement around the world for enterprise customers’ data, which it puts at “well under 1%” of the demands it receives. In 94 it was compelled to provide data: content in 45 cases, only non-content data such as subscriber information in 49, and “None of these disclosures involved Azure content data belonging to a commercial, public sector, or educational customer” [[8]]. These are requests for enterprise data in general, not for BitLocker keys.
The Structural Problem
The BitLocker story is one instance of a larger pattern. It is not that Microsoft behaved unusually or maliciously — it complied with a lawful order in its home jurisdiction, as it is legally required to do. The problem is structural: when an institution depends on a US-headquartered provider for key custody or critical hosted infrastructure, some operational control is shared with an entity whose legal obligations may conflict with the institution’s. Closed source can impede audit and exit, but provider control and jurisdiction — not the licence alone — create the production-order route.
This applies beyond encryption. It applies to email (Exchange Online, Outlook), document storage (SharePoint, OneDrive), communication (Teams), identity management (Entra ID, formerly Azure Active Directory), and any service that runs through a Microsoft account or Azure tenant. For each of these: the data is subject to Microsoft’s terms, and Microsoft is subject to US law.
The same argument applies, with different specifics, to Google Workspace and any other US-headquartered platform. The issue is not that these companies are bad actors. It is that their legal accountability and the legal accountability of European public institutions point in incompatible directions, and the institutions mostly have not noticed.
What Sovereign Software Looks Like
The alternative is not paranoia and air-gapped servers. It is a coherent strategy for institutional digital infrastructure that is based on software the institution controls.
In Germany, this conversation has a name and a project. openDesk bundles open-source tools into an integrated workspace alternative to Microsoft 365. As of October 2026 its documentation lists Nextcloud for files, Collabora Online for documents, OX App Suite for mail and calendars, Element on Matrix for chat, Jitsi for video calls, OpenProject, XWiki, CryptPad and Notes, with Univention’s Nubus, which uses Keycloak for single sign-on, for identity [[13]] [[14]]. Since January 2024 it has been developed by ZenDiS (Zentrum für Digitale Souveränität der Öffentlichen Verwaltung, the Centre for Digital Sovereignty of Public Administration), a company the federal government set up through the Federal Ministry of the Interior in 2022 and currently owns in full; the states’ participation is being prepared [[15]] [[16]]. The suite grew out of the ministry’s “Sovereign Workplace” initiative [[16]], the Souveräner Arbeitsplatz, which addresses the control problem that the BitLocker story illustrates. Open source permits audit and self-hosting; it does not by itself guarantee that keys remain local or remove every foreign legal dependency. Deployment, operator, identity provider, support contracts, and key management decide that.
Public bodies have begun to use it. According to ZenDiS, the Conference of Minister-Presidents used openDesk at its October 2024 meeting in Leipzig for documenting resolutions and exchanging and co-editing files among the sixteen states [[17]]. In 2025 the Bundeswehr’s IT provider BWI signed a framework agreement with ZenDiS, and the Robert Koch Institute contracted to bring openDesk to about 7,000 users of the public health service’s Agora platform [[18]]. Schleswig-Holstein has gone further with its own open-source stack: in October 2025 it finished moving more than 40,000 mailboxes of its state administration from Exchange and Outlook to Open-Xchange and Thunderbird [[19]].
The city of Munich’s earlier experiment with Linux (LiMux) and its partial rollback to Microsoft products is a cautionary tale, but not a controlled test of open-source fitness. Public accounts attribute the reversal to a mixture of technical, organisational, political, and vendor factors; reducing it to one failed technology or one lobbying campaign outruns the record. The BitLocker story is a reminder of what is at stake in that political negotiation.
The “Public Money? Public Code!” campaign of the Free Software Foundation Europe (FSFE), launched in 2017 with an open letter from 31 organisations [[20]], has articulated the principle cleanly: the letter asks for legislation “requiring that publicly financed software developed for public sector must be made publicly available under a Free and Open Source Software licence” [[21]]. The argument is not only about freedom as an abstract value. It is about the practical consequence of being locked into a proprietary platform: your institution loses the ability to audit what the software does, to modify it to meet your requirements, to host it where your data protection law applies, and to switch providers without losing access to your own data.
What I Do, and Why
I work at a publicly funded institution. The software I build for institutional contexts — campus infrastructure, workforce management, archival systems, alert systems — is public.
Not because I am ideologically committed to open source as a movement, but because the alternative is incoherent. If I build tooling for a university with public funds and keep it closed, I have produced a private asset with public money, duplicated by every institution that builds the same thing independently, inspectable by nobody, and ultimately dependent on my continued willingness to maintain it or hand it over. None of those outcomes serve the institutions I am building for.
Here is what that looks like in practice:
chronikwerk (since renamed from zammad-ticket-archiver) — archival of Zammad support tickets as PDFs with adjacent JSON audit records, triggered by authenticated webhooks. It can optionally apply a PAdES signature and an RFC 3161 timestamp. It is a self-hosted service with filesystem storage and an alpha-candidate release status; the operator supplies any signing material.
escalane (since renamed from alarm-broker) — a self-hosted alarm service that turns device triggers (currently Yealink-compatible HTTP) into an operator worklist and acknowledgement links for responders. It records each alarm’s history in PostgreSQL and uses background workers to send notifications through Zammad, SendXMS (SMS) and a Signal REST bridge, and to escalate unanswered alarms. It is a public alpha and has not been validated for emergency response.
concourse (since renamed from campus-app-kit) — a TypeScript workspace for presenting public university information: a Node.js API reads an institution pack, normalises public campus web pages and ICS feeds, and serves validated JSON to an Expo client for native and web. It covers events, rooms defined in the pack, and schedules. It deliberately includes no protected connectors, credentials, SSO or user accounts.
cueq — an integrated workforce management system for German universities under TV-L (the collective agreement for public sector employees in the German states). Handles time recording, shift planning, absence management, monthly closing, exports, and audit records. Built around NestJS and Next.js, with a PostgreSQL backend and Honeywell terminal integration, and designed so that HR data stays on the institution’s own infrastructure. It is pre-release source for evaluation with synthetic data, not yet a production system for real employee records.
These are all boring. They are not research contributions; they are plumbing. But plumbing is what holds institutions together, and the question of who controls the plumbing — and under whose legal jurisdiction — is exactly the question the BitLocker story makes visible.
The Principle
Public money, public code. If an institution funded by public money develops software for its own operations, that software should be released under an open licence, inspectable, forkable, and deployable by any institution with the same needs.
The corollary: institutions funded by public money should prefer software that is itself openly licensed, auditable, and deployable on infrastructure the institution controls. Not as a blanket ban on proprietary tools where they are genuinely the best option, but as a starting presumption that shifts the burden of justification.
The BitLocker story is not a story about Microsoft doing something wrong. It is a story about the logical consequence of a procurement decision that was made without asking “and what happens when a US court sends a warrant?” That question was available in 2018 when the CLOUD Act passed, in 2020 when Schrems II was decided, and before both. It is still available now, for every institution that has not yet asked it.
Legal, policy and product documentation checked through 2026-10-28.
The FSFE’s “Public Money? Public Code!” campaign is at publiccode.eu; openDesk is at opendesk.eu.
References
[1] Brewster, T. (2026, January 22). [Forbes report on Microsoft and BitLocker keys; headline not seen]. Forbes. https://www.forbes.com/sites/thomasbrewster/2026/01/22/microsoft-gave-fbi-keys-to-unlock-bitlocker-encrypted-data/ Not opened for this revision (the page refuses automated access). Green’s words are quoted via Holwerda, T. (2026, January 23), OSnews, reposted on Pipedot: https://pipedot.org/732BM
[2] Franceschi-Bicchierai, L. (2026, January 23). Microsoft gave FBI a set of BitLocker encryption keys to unlock suspects’ laptops: Reports. TechCrunch. https://techcrunch.com/2026/01/23/microsoft-gave-fbi-a-set-of-bitlocker-encryption-keys-to-unlock-suspects-laptops-reports/
[3] Microsoft. Back up your BitLocker recovery key. Microsoft Support. Retrieved 28 October 2026 from https://support.microsoft.com/en-US/Windows/Security/encryption/back-up-your-bitlocker-recovery-key
[4] Microsoft. BitLocker overview (section “Device encryption”). Microsoft Learn, 24 August 2026. Retrieved 28 October 2026 from https://learn.microsoft.com/en-us/windows/security/operating-system-security/data-protection/bitlocker/
[5] Microsoft. BitLocker recovery overview (section “BitLocker recovery password”). Microsoft Learn, 29 July 2025. Retrieved 28 October 2026 from https://learn.microsoft.com/en-us/windows/security/operating-system-security/data-protection/bitlocker/recovery-overview
[6] Mossdorf, K.-S. (2026, January 26). BitLocker: Microsoft hands over keys to law enforcement. heise online. https://www.heise.de/en/news/BitLocker-Microsoft-provides-keys-to-law-enforcement-11153016.html
[7] 18 U.S.C. § 2713. Required preservation and disclosure of communications and records. Added by Pub. L. 115–141, div. V, § 103(a)(1) (the CLOUD Act), 23 March 2018, 132 Stat. 1214. Retrieved 28 October 2026 from https://www.law.cornell.edu/uscode/text/18/2713
[8] Microsoft. Law Enforcement Requests Report: requests for enterprise customer data, July–December 2025. Retrieved 28 October 2026 from https://www.microsoft.com/en-us/corporate-responsibility/reports/government-requests/customer-data
[9] Microsoft. Microsoft Privacy Statement (section “Storage and processing of personal data”), last updated September 2026. Retrieved 28 October 2026 from https://www.microsoft.com/en-us/privacy/privacystatement
[10] Court of Justice of the European Union. (2020, July 16). Press release No 91/20: Judgment in Case C-311/18, Data Protection Commissioner v Facebook Ireland and Maximillian Schrems. https://curia.europa.eu/jcms/upload/docs/application/pdf/2020-07/cp200091en.pdf
[11] European Commission. EU-US data transfers (adequacy decision for the EU-US Data Privacy Framework, adopted 10 July 2023). Retrieved 28 October 2026 from https://commission.europa.eu/law/law-topic/data-protection/international-dimension-data-protection/eu-us-data-transfers_en
[12] Appeal brought on 31 October 2025 by Philippe Latombe against the judgment of the General Court (Tenth Chamber, Extended Composition) delivered on 3 September 2025 in Case T-553/23, Latombe v Commission (Case C-703/25 P). Official Journal of the European Union, C series, C/2025/6610, 22 December 2025. https://data.europa.eu/eli/C/2025/6610/oj
[13] ZenDiS. openDesk documentation: Introduction (functional components). Retrieved 28 October 2026 from https://docs.opendesk.eu/operations/introduction
[14] Univention. Nubus for Kubernetes: Architecture manual, section 3.4, Identity Provider. Retrieved 28 October 2026 from https://docs.software-univention.de/nubus-kubernetes-architecture/1.x/en/components/identity-provider.html
[15] ZenDiS. Wer wir sind. Retrieved 28 October 2026 from https://www.zendis.de/wer-wir-sind
[16] ZenDiS. openDesk: The office and collaboration suite for public administration. Retrieved 28 October 2026 from https://www.opendesk.eu/en/
[17] Niemeyer, L. (2025, October 23). Use case: MPK relies on openDesk for secure collaboration. openDesk blog (ZenDiS). https://www.opendesk.eu/en/blog/digital-collaboration-public-administration
[18] Krempl, S. (2025, June 13). Sovereignty: Public health service increasingly relies on open source. heise online. https://heise.de/-10443654
[19] Krempl, S. (2025, October 7). Schleswig-Holstein’s e-mail systems converted to open source. heise online. https://heise.de/-10733737
[20] Free Software Foundation Europe. (2017, September 13). Public Money? Public Code! 31 organisations ask to improve public procurement of software. https://fsfe.org/news/2017/news-20170913-01.en.html
[21] Free Software Foundation Europe. Public Money? Public Code! Open letter. Retrieved 28 October 2026 from https://publiccode.eu/en/openletter/
Changelog
- 2026-07-11: Corrected the Windows recovery-key wording: eligible devices can back a key up to a Microsoft account, but the exact behaviour depends on device, account and organisation policy. Added formal-source review for the CLOUD Act, GDPR, Schrems II, OpenDesk and LiMux history.
- 2026-10-03: Corrected the attribution of the “if you have access to keys, eventually law enforcement is going to come” quote from Bruce Schneier to Matthew Green (Forbes, 2026-01-22; Schneier’s post only quoted Forbes on subpoenas), and credited Forbes, which first reported Microsoft’s confirmation. Replaced “Azure Active Directory” with its current name, Entra ID. Described OpenDesk’s developer as ZenDiS, founded by the federal government, rather than the federal and state governments jointly. Fixed the cueq repository link, which returned 404 (the repository is
cueq). Refreshed the project descriptions against the current repositories: zammad-ticket-archiver is now chronikwerk, alarm-broker is now escalane and campus-app-kit is now concourse, with feature descriptions corrected (for example, concourse has no room booking or private connectors, and the unverified self-hosted-keys and no-external-dependency claims were removed). Changed the summary’s “this week” to “last week” to match the body. Aligned the cueq description with its repository: pre-release source for evaluation with synthetic data; “payroll export” and “GDPR-compliant audit trails” replaced by the documented exports and audit records. - 2026-10-28: Replaced the quotation “quite dangerous” attributed to Jennifer Granick, which I could not check (Forbes refuses automated access and no reachable report quotes those words), with the one statement heise reports from her: that this data is likely of great interest to many governments. The point that other governments can use the same route is now marked as my reading. Corrected “the original TechCrunch reporting”: Forbes broke the story and TechCrunch followed. The handover is now given as Forbes’s report rather than as Microsoft’s confirmation, in the summary as well: Microsoft told Forbes that it sometimes provides such keys and receives about 20 requests a year, and the warrant comes from local Guam reporting. Changed “subpoena” to “warrant” in the closing question. Added how device encryption backs the key up (Entra ID, Active Directory, or an administrator’s Microsoft account, which Microsoft calls the default for devices joined to neither) and the text of 18 U.S.C. § 2713. Added that the reports do not say where the Guam keys were stored, that I found no Microsoft statement of where Microsoft-account recovery keys sit, and what Microsoft’s privacy statement says about where personal data is stored. Added the 2023 EU-US Data Privacy Framework and the pending appeal in Latombe. Added Microsoft’s policy of redirecting enterprise requests to the customer and notifying the customer, with its figures for the second half of 2025 for enterprise data in general, not BitLocker keys: data disclosed for 94 of 190 requests, and content data of three non-US customers stored outside the US given to US law enforcement. Replaced “several German states and federal agencies have been piloting OpenDesk” with documented cases (the Minister-Presidents’ conference of October 2024 according to ZenDiS, BWI and the Robert Koch Institute in 2025, and Schleswig-Holstein’s mail migration). Brought the component list in line with the openDesk documentation as of October 2026 (OX App Suite, OpenProject, XWiki, CryptPad and Notes added; identity through Nubus, which uses Keycloak for single sign-on). Corrected the gloss of Souveräner Arbeitsplatz from “sovereign workspace” to “sovereign workplace”, spelled the product openDesk as ZenDiS does, and quoted the FSFE campaign’s 2017 open letter. Converted the source list to numbered references; the uscode.house.gov link returned “Page Not Found” when checked and now points to Cornell’s LII copy, and the openDesk link points to opendesk.eu. Added Fig. 1.