AI Meeting Note-Takers for Nordic Businesses (2026): Who Actually Keeps Your Recordings in the EU?

The tool that records everyone in the room – AI Meeting Note-Takers for Nordic Businesses (2026)

You switch on an AI note-taker for a client call. It transcribes everything, writes a tidy summary, and saves you twenty minutes. But think about what just happened to the data: a recording of your client’s voice, their words, maybe commercially sensitive or personal details, has been captured and sent to a server somewhere. Your client never agreed to that. Neither did the candidate in your interview, or the partner on the other call.

That is what makes meeting note-takers different from most AI tools. A recording captures the personal data of everyone present, under GDPR, and where that data goes is not a footnote, it is the whole question. So this guide is not about which tool writes the nicest summary. It is about which ones keep your recordings in the EU, which train their AI on them, which quietly send everything to US servers, and which one is actually built in the Nordics.

The honest shortlist:

• Want a Nordic-built option? Proudfrog (Sweden) is the only one we found, and it stores data in Finland. Genuinely promising, with some real compliance gaps to know about.
• Want a clean EU-hosted tool with the paperwork sorted? tl;dv and Jamie (both German), Leexi (Belgian), and HappyScribe (Irish) all keep data in the EU by default.
• Worried about a recording bot joining every call? Proudfrog, Jamie and Hedy record on your own device, no bot, no one else sees it happening.
• Considering a big US name (Otter, Fathom, Fireflies)? Know that most store and process in the US, and two of them train their AI on your recordings by default.

The distinction that matters most: storage versus processing

Here is the thing almost every vendor blurs, and it is the key to reading this whole category honestly.

“Your data is stored in the EU” and “your data is processed in the EU” are not the same claim. A tool can keep your saved transcripts on an EU server while sending the actual audio to a US company for transcription, and to another US company for the AI summary. During that processing, your meeting leaves the EU, even though the stored file sits in Frankfurt or Helsinki.

So when a tool says “EU data residency,” the real questions are: does that cover processing or only storage? And is it the default, or only on the expensive Enterprise plan? Several tools that market themselves as EU-friendly turn out to store in the EU but process through US sub-processors, or to offer EU residency only if you pay for Enterprise. We flag which is which below.

The Nordic-built option: Proudfrog 🇸🇪

🔵 Verified from documentation.

Proudfrog is a Swedish company (Proudfrog AB, Stockholm), and it is the only genuinely Nordic-built meeting note-taker we could find. It gets several things right that a Nordic business will care about. It records without a bot (through iOS and macOS apps that capture on your device, so nothing joins the call), it confirms accurate transcripts in all four Nordic languages (Swedish, Norwegian, Danish and Finnish) with speaker identification, and it stores data in the EU. Its pricing is refreshingly simple too: no subscription, pay-per-meeting from around €0.36 per hour.

On storage, there is a small detail worth getting right. The homepage says “data stays in Sweden,” while the privacy policy specifies Google Cloud’s europe-north1 region, which is physically in Finland. Either way the data stays in the Nordics and the EU, but the precise location is Finland, not Sweden. A Nordic-built tool keeping Nordic meeting data on Nordic soil is a real data-sovereignty story, and one no US competitor can tell.

But we have to be straight about the caveats, because they matter for a business:

• Storage is in the EU, but processing is not. The audio is sent to US sub-processors for transcription (AssemblyAI, OpenAI) and to US AI providers for summaries (Anthropic, xAI). So while your stored transcripts stay in the EU, the recording itself is processed in the US. The “data stays in Sweden” message refers to storage, not the full journey.
• No published DPA. We could not find a Data Processing Agreement anywhere on the site. A business using Proudfrog as a data processor needs a signed DPA under GDPR Article 28, and it is not currently offered.
• No documented transfer mechanism. The privacy policy names US sub-processors but does not state the legal basis for those transfers (no SCCs, no Data Privacy Framework reference). That is a genuine gap in the compliance paperwork, despite the “GDPR compliant” badge on the homepage.

The honest verdict: the most Nordic option by a distance, with genuinely good instincts, bot-free, all four Nordic languages, EU storage, and fair pricing. But it reads as an early-stage startup that has not yet built the enterprise compliance layer (DPA, documented transfers) that a larger Nordic business would require. Great for a small Nordic team comfortable with the trade-offs; not yet ready for a compliance-strict buyer. Genuinely worth watching as it matures. Verdict: 🟡 The Nordic champion, promising but not yet enterprise-ready on paperwork.

The clean EU-hosted tools

These keep data in the EU by default and have the compliance documentation a business actually needs.

tl;dv 🇩🇪 (and its Finland servers)

🔵 Verified from documentation. A German company (Aachen), tl;dv stores data in the EEA by default for all users, including free. For Nordic readers there is a standout detail: part of its infrastructure runs on Hetzner servers in Finland, so Finnish meeting data may sit in Finland. It also does something stronger than most: it contractually prohibits its AI sub-processors (including Anthropic and Google) from training on your content, not just itself. Data does not train models, by contract, all the way down the chain.

Watch for: its DPA page was inaccessible at the time of research (confirm it directly), and AI processing can optionally use US inference, so choose the EU option. Verdict: 🟢 Strong EU-default choice, with a genuine Finland-hosting angle.

Jamie 🇩🇪 (the bot-free one)

🔵 Verified from documentation. A German company, Jamie’s differentiator is that no bot joins your call. It runs as a desktop app, records audio locally on your device, and no one else in the meeting sees a recording bot appear. Meeting content is stored in Google Cloud Frankfurt, audio is deleted immediately after transcription, and it never trains on your content. A clean privacy story.

One important caveat to know: Jamie has a “Speaker Memory” feature that builds voice profiles to recognise speakers across meetings. That is biometric data under GDPR Article 9, and it requires explicit consent from every participant. Jamie does not provide tooling to collect that consent, so the obligation sits with you. Leave the feature off unless you have handled consent properly. Verdict: 🟢 Best bot-free EU option, with a biometric feature to use carefully.

HappyScribe 🇮🇪

🔵 Verified from documentation. An Irish company, EU data storage by default on all plans, and it confirms support for all four Nordic languages (Finnish, Swedish, Norwegian Bokmål, and Danish, though not Nynorsk). Training on your content is opt-in only, so off by default. A solid, straightforward EU choice with the broadest confirmed Nordic-language coverage here.

Watch for: the primary product uses a recording bot (a mobile app offers bot-free recording), and no per-language accuracy figures are published, test Finnish on your own audio. Verdict: 🟢 Reliable EU-default option, strongest confirmed Nordic-language support.

Leexi 🇧🇪

🔵 Verified from documentation. A Belgian company hosting exclusively in the EU, Leexi stands out on compliance credentials: it publishes its DPA openly (no sales call needed), uses EU-based AI sub-processors (Gladia and Mistral, both EU), and holds ISO 42001 certification (the AI-management-system standard), which is genuine EU AI Act preparedness. If documented compliance is your priority, Leexi has the strongest paperwork of the group.

Watch for: confirm Nordic-language quality on your own audio; its focus leans toward sales-conversation intelligence. Verdict: 🟢 The strongest compliance credentials, EU-only hosting.

Hedy 🇺🇸/🇪🇺 (US company, EU option)

🔵 Verified from documentation. Worth including for one reason: Hedy is a US company (Portland, Oregon), but it offers EU data residency (Google Cloud, Belgium/Netherlands) that you select at signup, and it processes audio on your device where the hardware supports it, so audio need not leave your machine at all. A US entity with a genuinely privacy-forward architecture.

Watch for: it is still a US-headquartered company (the US-parent question applies), and crucially, it publishes no language list, so Finnish and other Nordic support is unconfirmed. Verify before relying on it for Nordic-language meetings. Verdict: 🟡 Privacy-forward and EU-selectable, but US parent and unconfirmed Nordic languages.

The US incumbents: know what you are signing up for

These are the famous names, and they are capable products. But for a Nordic business handling meeting recordings, each carries a real caveat.

Otter

🔵 Verified from documentation. A US company (Delaware), all infrastructure US-based, no EU data residency at any tier. More importantly, Otter trains its own AI on your recordings and transcripts by default (its privacy policy states it trains on de-identified audio and transcriptions, and it uses human labellers). It also processes voiceprints to identify speakers, which is biometric data. For a Nordic business, this is the hardest of the group to justify with client or candidate recordings. Verdict: 🔴 US-only, trains on your recordings by default. Avoid for sensitive Nordic meetings.

Fathom

🔵 Verified from documentation. A US company (San Francisco), hosted in the US, no EU residency option found at any tier. It trains its in-house AI on your meeting content by default, though it does offer an opt-out in account settings (available on all plans, so turn it off). To its credit, it contractually stops third parties (OpenAI, Anthropic, Google) from training on your data, the training is in-house only. Still US-hosted throughout. Verdict: 🔴 US-only hosting; trains by default but with an opt-out to set.

Fireflies

🔵 Verified from documentation. A US company, US storage and processing by default. EU storage exists but only on the Enterprise tier (via “Private Storage”), and even then EU processing is described as a future capability, not yet available. On the positive side, Fireflies does not train on your content and contractually prohibits its vendors from doing so. Two things to note for compliance: its privacy policy carries defensive biometric-law language around voice data (it treats voice as potential biometric identifiers under Illinois, Texas and Washington law), and it offers a sentiment-analysis feature, which is worth flagging because sentiment or emotion analysis applied in HR or employment contexts can fall into EU AI Act high-risk territory. Verdict: 🟡 Does not train on your data, but EU storage is Enterprise-only and EU processing is not yet available.

At a glance

Tool	Origin	EU data residency	Trains on your content?	Bot-free option?	Verdict
Proudfrog	Sweden 🇸🇪	Storage in Finland (processing US)	No	Yes (on-device)	🟡
tl;dv	Germany 🇩🇪	Default (Finland servers)	No	No	🟢
Jamie	Germany 🇩🇪	Default (Frankfurt)	No	Yes (on-device)	🟢
HappyScribe	Ireland 🇮🇪	Default	Opt-in only	Via mobile app	🟢
Leexi	Belgium 🇧🇪	EU-only, ISO 42001	No	No	🟢
Hedy	US 🇺🇸	Selectable + on-device	No	Yes (on-device)	🟡
Otter	US 🇺🇸	None	Yes, by default	No	🔴
Fathom	US 🇺🇸	None	Yes (opt-out)	No	🔴
Fireflies	US 🇺🇸	Enterprise-only (storage)	No	No	🟡

Verified from each vendor’s own documentation (🔵), not hands-on tested. Nordic-language quality and current terms should be confirmed on your own audio and the vendor’s live pages before committing.

The consent problem nobody mentions

Whichever tool you pick, there is a duty that sits with you, not the vendor: the other people in the meeting have rights too. A recording captures their personal data, and under GDPR you generally need a lawful basis and, in practice, to tell them.

Two things follow. First, a bot-free tool (Proudfrog, Jamie, Hedy) that records on your device is quieter, but that is a double edge: because no bot appears in the participant list, no one else knows recording is happening, which raises your obligation to tell them, not lowers it. Second, almost none of these tools provide consent-collection tooling, they place that duty on you in their terms. So build a simple habit of announcing the recording at the start of the call. It is good manners and good compliance.

How this connects to your wider compliance

Meeting recordings are one more place where “where does the data go” decides everything, the same question as every other AI tool. Our EU data residency guide covers that across the board. And the sentiment-analysis and voice-biometric features flagged above tie directly to our EU AI Act guide, which explains which AI uses count as high-risk.

Common questions

Which AI note-taker keeps data in the EU? For default EU hosting with the paperwork sorted: tl;dv, Jamie, Leexi and HappyScribe. Proudfrog (Swedish) stores in Finland but has compliance gaps. The big US tools (Otter, Fathom) offer no EU residency; Fireflies offers EU storage only on Enterprise.

Is there a Nordic-built meeting note-taker? Yes, Proudfrog, a Swedish company that stores data in Finland. It is promising but early, with no published DPA yet, so weigh that if you need strict compliance.

Which note-takers train their AI on my meetings? Otter trains by default, and Fathom trains by default with an opt-out. Most EU-native tools (tl;dv, Jamie, Leexi, HappyScribe, Fireflies) do not train on your content.

What is the difference between EU storage and EU processing? Storage is where the saved file sits; processing is where the audio is actually transcribed and summarised. A tool can store in the EU while processing through US servers, so check both.

Do I need to tell people I am recording? In practice, yes. A recording captures everyone’s personal data under GDPR. Announce it at the start of the call, especially with bot-free tools where no recording indicator appears.

Sources

All claims are from each vendor’s own published documentation, checked 13 August 2026. Compliance terms change often, so confirm against the live pages before relying on them. Links marked (JS) were reported as JavaScript-rendered or inaccessible to automated checking and should be opened manually.

Proudfrog (Sweden; EU/Finland storage, US processing, no published DPA)

• Privacy Policy · Terms · Homepage

tl;dv (Germany; EEA default, Hetzner Finland servers, no training incl. sub-processors)

• Privacy Policy · Imprint

Jamie (Germany; Frankfurt storage, bot-free, Speaker Memory biometric feature)

• Privacy Policy · Security

HappyScribe (Ireland; EU default, all four Nordic languages, opt-in training)

• Privacy Policy · Languages · Trust/sub-processors (JS)

Leexi (Belgium; EU-only hosting, ISO 42001, open DPA)

• Trust Center · Privacy Policy · DPA (PDF)

Hedy (US entity; EU-selectable residency, on-device processing)

• Privacy Policy · Security · Sub-processors (JS)

Otter (US; no EU residency, trains by default, voiceprints)

• Privacy Policy · Terms · Sub-processors

Fathom (US; no EU residency, trains by default with opt-out)

• Privacy Policy · DPA · Terms

Fireflies (US; EU storage Enterprise-only, no training, sentiment analysis)

• Data Processing Agreement · Privacy Policy · Security FAQ · Sub-processors (JS)

Note on lawsuits: this article does not assert any active lawsuit. Statements about biometric or voiceprint processing and sentiment-analysis features are taken from the vendors’ own privacy policies and feature pages, cited above. Any reference to specific litigation would require a separate court or news source and has been deliberately left out.

Keeping this up to date

This is a fast-moving category, new tools, new EU-residency options, and shifting terms. Everything here is from each vendor’s current documentation and dated. Last reviewed 18th of August 2026. If something has changed, tell us.

Published by NordicAITools Editorial. Assessments are based on each vendor’s published documentation (🔵), not hands-on testing. Compliance and language claims should be verified against vendors’ live terms before you commit, in line with our published methodology.