In early July 2026, the city council of Werne, North Rhine-Westphalia, voted unanimously to launch a pilot that will use AI transcription software to convert public portions of its council sessions into draft minutes. The six-month test explicitly prioritizes data protection, making Werne one of the first local governments to put privacy safeguards at the center of an AI meeting-tools rollout.

What Werne Actually Announced

According to the city’s own release, the pilot will see an AI-powered transcription service—widely expected to be the Microsoft Teams transcription engine, given its prevalence in German public-sector IT—process audio from the public segments of council meetings. The software will generate a raw text transcript, which clerks will then review and edit into formal minutes. Closed sessions, where confidential personnel or legal matters are discussed, will never be recorded or transcribed.

The council included two key data-protection conditions: transcripts will be stored exclusively on servers within the European Union, and personally identifiable information will be scrubbed from the AI’s output before any human review. The project runs through December 2026, after which a full privacy audit and accuracy assessment will determine whether to continue.

Why Data Protection Took the Spotlight

Werne’s council didn’t stumble into this decision. German data-protection authorities have long scrutinized AI transcription, especially when it involves public bodies. The General Data Protection Regulation (GDPR) treats voice recordings as personal data, and any processing by an AI model requires a lawful basis—typically the explicit consent of all participants or a legitimate public interest that passes a strict necessity test.

For a city council meeting, attendees include elected officials, city staff, and sometimes members of the public who speak during open-comment periods. Recording their voices and running them through a cloud AI immediately raises questions about storage, access, and retention. Werne’s decision to limit transcription to the public portion and to anonymize the draft transcripts addresses those concerns head-on. It also sidesteps the thornier issue of obtaining 100 percent consent from a roomful of people, a challenge that has tripped up several early government AI pilots.

What This Means for Microsoft Teams Users

Even if you aren’t a German civil servant, Werne’s pilot matters. Microsoft Teams is the default collaboration hub for millions of Windows users, and its live transcription feature is available in any scheduled meeting. The same AI engine that will turn council debates into draft minutes is the one you might use to capture a project standup or a client call.

Here’s what every Teams user should understand:

  • Transcription is not the same as recording. When you turn on live transcription, Teams displays real-time captions and saves a text transcript after the meeting. That transcript can include speaker names and timestamps. Your IT admin can control whether it’s automatically generated or requires manual triggering.
  • Data residency matters. In Werne’s case, the council insisted on EU-based storage. For your organization, Teams transcription data is stored in the region associated with the meeting organizer’s Microsoft 365 tenant—unless you have purchased the Multi-Geo add-on and configured it. Check with your admin to know where your meeting words land.
  • Consent prompts are optional. Admins can require a consent banner that asks all participants to agree before transcription starts. Without it, one click from the organizer can record everyone’s words into a file that persists. For privacy-sensitive meetings, demand that your admin enables the consent prompt (found in the Teams admin center under Meetings > Meeting policies).

A Quick Refresher: How Teams AI Transcription Works

Microsoft introduced live transcription in Teams back in 2020, initially for scheduled meetings only. It uses automatic speech recognition (ASR) trained on general conversational data. The service runs in the cloud, processing audio streams from all participants in real time. The resulting transcript is stored alongside the meeting metadata and can be deleted manually or via retention policies.

Crucially, Microsoft does not use your Teams transcription data to train its AI models, a point the company has reinforced in its privacy documentation following multiple GDPR inquiries. Yet the service still sends audio to Microsoft servers, which means government users must conduct a data protection impact assessment (DPIA) before turning it on. Werne’s pilot effectively serves as a public DPIA, testing whether the technical and organizational measures it outlined actually work.

What Governments and Enterprises Need to Do Now

Werne’s cautious approach offers a blueprint for any organization, public or private, considering AI meeting tools:

  1. Define exactly what gets transcribed. Separate public and confidential sessions; in Teams, that might mean separate meetings or a clear agenda point where transcription is paused.
  2. Configure data residency. Verify that your Teams tenant stores transcripts in an approved geography. For EU entities, Microsoft offers the “EU Data Boundary” that aims to keep customer data within the EU, but transcription may still be processed in the US unless you have add-ons. Read the fine print.
  3. Enable the consent prompt. In the Teams admin center, go to Meeting policies, select your policy, and under “Recording & transcription,” turn on “Auto-recording consent.” This forces organizers to get a click-through from attendees before any transcript is created.
  4. Set a short retention period. By default, Teams transcripts are kept indefinitely. Use Microsoft Purview retention policies to auto-delete transcripts after 30, 60, or 90 days, unless a legal hold applies. Werne’s pilot, for instance, will delete all transcripts after the audit phase.
  5. Audit AI accuracy and bias. Council meetings often involve local dialects, overlapping speech, and background noise. A transparency report from the pilot should measure how often the AI misattributes speakers or mishears terms. Your organization should do the same: spot-check transcripts against manual notes to catch systematic errors.

For individual users, the lesson is simpler: when you see the transcription icon light up red in a Teams meeting, your speech is being captured. If you’re discussing sensitive topics, ask the organizer to turn it off—or, at least, to use the consent banner so everyone is warned.

The Bigger Picture: AI in Civic Life

Werne’s pilot is small in scale but significant in signal. Across Europe and North America, local governments are experimenting with AI for everything from summarizing public comments to drafting legislation. The European Commission’s AI Office is currently drafting a code of practice under the AI Act, and meeting transcription will likely fall into the “limited risk” category, requiring transparency obligations.

Microsoft, meanwhile, continues to bake Copilot and AI summarization into Teams. Later this year, Windows 11 is expected to gain deeper AI meeting recall features that mesh with Team’s transcripts. For administrators, staying ahead means mastering the privacy controls now, before the tools become ubiquitous.

As for Werne, the city expects to publish its findings in early 2027. If the numbers show that AI saves clerks’ hours without compromising privacy, the pilot could become permanent—and a template for thousands of other councils weighing the same choice.