Fathom's free plan gives you unlimited recordings and transcriptions at no cost. That generosity is real, and it is also the question. Fathom stores all of your meeting data in the United States, sends it to US AI vendors to process, and by default uses de-identified content from your meetings to train its own AI. For a European team, "free" is priced in where your data goes, not in euros.
Here is what Fathom's own documents say about the data path, why the product can afford to be free, and what an EU-sovereign alternative looks like.
Where Fathom stores your meeting data
Fathom is direct about this in its own documentation. The privacy policy states that its services are "hosted in the United States," and its help centre says plainly that "all Fathom data is stored in the United States." Data is encrypted at rest and in transit, which is table stakes and not the point.
The point is what is missing. Fathom does not offer an EU or EEA data-residency option. There is no European region you can pick. For a European customer, the practical effect is the same as it is with most US notetakers: your meeting audio and transcripts are transferred to and held in the US, and EU customers get contractual protection rather than data residency.
That contractual protection is the EU-US Data Privacy Framework, under which Fathom self-certifies, backed by Standard Contractual Clauses as a fallback if the framework lapses. Its Data Processing Agreement, effective May 2026, spells this out. Those are valid transfer mechanisms. What they are not is a promise that the data stays in Europe. They exist to legitimise moving it to the US, where it becomes reachable under US law.
Does Fathom train its AI on your conversations?
This is where precision matters, because Fathom's answer is "yes, our own AI, by default, but not third parties."
Fathom's privacy policy states that, based on your account settings, it "may use and create de-identified data generated from Meeting Content Information to improve our Services by training, improving, and customizing our in-house artificial intelligence models." It adds that you can opt out in your account settings. So the default is on, and Fathom's own model improvement runs on de-identified content taken from your meetings unless you go and switch it off.
Separately, Fathom is clear that it does not let its third-party AI providers train on the data. It names those providers as Anthropic, OpenAI, and Google, and its help centre describes closed-API arrangements that "restrict either company from retaining, accessing or training on Fathom or customer data." Both things are true at once: third-party vendors are contractually restricted, while Fathom's own model training uses de-identified material by default.
De-identification is worth less than it sounds. Regulators increasingly reject the idea that stripping obvious identifiers anonymises voice and conversation data. If your meetings carry client names, deal terms, or personal circumstances, "de-identified" is carrying a lot of weight, and it is on by default.
Your meeting content still travels to US AI vendors
Even setting training aside, the processing path matters. Fathom sends meeting content to US AI subprocessors, named in its own documentation as Anthropic, OpenAI, and Google, to generate summaries and answers. The contractual no-training terms are a genuine protection, but they do not change the jurisdiction. Your conversation is leaving for US-owned AI APIs to be processed, and it is stored on US infrastructure the whole time.
That is a structurally different position from a provider that never moves the data out of the EU. Contractual promises depend on the other party keeping them and on the transfer framework surviving the next legal challenge. Residency does not.
So why is Fathom free?
Free is a strategy, not a gift. Fathom is a US venture-backed company, founded in 2020 by Richard White, the founder of UserVoice. It raised a 4.7 million dollar seed round in 2022, backed by early Zoom investors and Zoom's own apps fund, and a 17 million dollar Series A led by Telescope Partners in September 2024, roughly 21.7 million dollars in total.
Its founder has been open about the plan. In interviews he has described a deliberate choice to stay free for the first couple of years and win user attention before monetising advanced features, on the logic that a business you can build without making money for a year or two is hard for competitors to catch. The free tier is a distribution engine funded by investors and paid upgrades, and the way it wins is by getting into as many meetings as possible.
None of that is sinister. It is a normal venture playbook. But it does answer the headline question: the free plan is valuable to Fathom precisely because of the volume of meeting data flowing through it, and that data flows to US infrastructure and US AI vendors, with de-identified content feeding Fathom's own models by default. We wrote about this pattern across the category in the hidden cost of free AI notetakers: the price of a free notetaker is usually paid in data exposure.
The free plan carries the full exposure
If you are on Fathom's free tier because "it is only notes," the data posture does not soften. The same US storage, the same US AI subprocessors, the same Data Privacy Framework and SCC transfer model, and the same default-on training on de-identified content apply regardless of tier. Paying moves you up the feature ladder; it does not move your data into Europe, because there is no European option to move it into.
The EU alternative
If the requirement is that meeting data stays under EU jurisdiction, the fix is not a more careful US vendor, it is a sovereign one. Numi's meeting assistant keeps recording, transcription, storage, and analysis inside EU jurisdiction on infrastructure you can point to. No US processing. No training on your data. The meeting bot joins Microsoft Teams and Google Meet through an EU-only pipeline, and the integrations fit how your team already meets.
If you are weighing tools side by side, our GDPR-compliant AI meeting assistant comparison scores the leading options against the criteria that actually decide compliance: data residency for storage and processing, Article 28 DPA terms, transfer impact assessments, and no-training obligations. And if you want the per-tool version of this investigation, we traced the same questions for Otter.ai and Fireflies.ai.
Fathom is a well-built product, and free is a strong offer. For a European team that answers to GDPR and to its own customers, the deciding question is not what it costs. It is where your meeting data lives, who processes it, and what it is used for once it gets there.