Why this matters more for an AI workspace than for a chatbot

When you ask a chatbot a question, you choose what to paste in. An AI workspace is different by design: to be useful, it holds your inbox, your calendar and your files on an ongoing basis. That is the source of its value — and the source of the obligation. Your work email is not just your personal data; it is full of clients', colleagues' and prospects' personal data, and under the GDPR you are responsible for how it is processed.

The stakes are not theoretical. Under the GDPR, the most serious infringements can draw fines of up to €20 million, or 4% of a company's total worldwide annual turnover, whichever is higher.1 Add the EU AI Act, which is phasing in its obligations across 2025–2027, and European teams have a straightforward incentive to get the tooling right the first time rather than retrofit it.

What "GDPR-compliant" actually requires

No vendor can hand you a single certificate that makes you compliant — compliance is about your whole processing setup. But a workspace can either make that easy or impossible. These are the properties to verify:

  • EU/EEA data residency. Where is your email, calendar and document data actually stored and processed? "We have EU customers" is not the same as "your data stays in the EU." Ask for the region.
  • A data processing agreement (DPA). Under Article 28, your vendor is your processor and must sign a DPA setting out what they may do with the data. No DPA, no compliant deployment for a business.
  • Transparent sub-processors. Most tools rely on others (hosting, email infrastructure, AI model providers). You are entitled to a current list of who processes your data and where.
  • No training on your content. Your emails and files must not become training data for someone's model. This should be a contractual commitment, not a setting buried in a menu.
  • Encryption and isolation. Stored credentials and content encrypted; strict per-account isolation so one customer can never reach another's data.
  • Data-subject rights. A workable path to access, export and deletion, so you can honour requests from the people whose data you hold.
  • A human in the loop. For consequential actions — sending, deleting, cancelling — the AI should prepare, and a person should approve. Automated decisions with legal or similarly significant effects are exactly what regulators scrutinise.

The quick test. Ask a vendor three questions: Which EU region is my data in? Will you sign a DPA? Do you train any model on my content? If the answers are vague, "it depends," or "we're working on it," you have your answer — whatever the feature list says.

Why most US AI assistants struggle here

This is not a knock on the quality of US products — many are excellent. It is about fit. A general-purpose US chatbot is not GDPR-compliant out of the box for business email: on consumer tiers there is typically no DPA, no EU data-residency commitment, and limited control over whether your inputs improve the model.2 You can sometimes reach a compliant setup through an enterprise agreement or an EU-hosted wrapper, but that is work you are doing to make a tool fit a jurisdiction it was not built for.

The newest wave of consumer "AI agents" — the personal assistants that read your inbox and act for you — often make this harder, not easier: several are US-only, launch without a European data-processing agreement, or default to training on user data. For a European business, that turns an otherwise impressive product into a compliance project. We went through the specific trade-offs in the European alternative to Muse, Instinct, Grok Bot and dots.

What the European-built version looks like

Usenti is an AI workspace designed for this constraint from the start rather than bolted onto it. Concretely:

  • A Danish company. Usenti I/S is incorporated in Denmark and supervised by the Danish data protection authority, Datatilsynet.
  • EU hosting. Data is hosted in the EU — the encrypted system that holds and moves it is called The Vault.
  • AES-256 and isolation. Stored email credentials and meeting transcripts are encrypted; row-level security isolates every account.
  • No training on your content. Customer content is never used to train AI models.
  • A DPA and sub-processor transparency. A data processing agreement for business customers, and a published list of who processes data and where.
  • Human approval on irreversible actions. Donna, the built-in assistant, drafts and prepares, but a person clicks before anything is sent, deleted or cancelled.

Usenti publishes its full security posture, including known limitations and current certification status, so you can check rather than take marketing's word for it. (A note on honesty, because it is the whole point of this article: being EU-hosted and supervised by a data authority is not the same as holding a SOC 2 or ISO 27001 certificate — ask every vendor, including this one, exactly which it has today.)

If you want the broader picture of what an AI workspace is before weighing the privacy angle, start with what an AI workspace actually is; if the cost of consolidating your stack is the deciding factor, see replacing your SaaS stack with one AI workspace.

A workspace that keeps your data in Europe

Usenti brings your email, calendar, meetings, docs and tasks into one EU-hosted system, with an AI assistant that works across all of it — Copenhagen-based, supervised by Datatilsynet, encrypted, and never trained on your content.

Get started for $1/month

$1/mo for 3 months + AI usage · then $45/user/mo · no free plan

Frequently asked questions

What makes an AI workspace GDPR-compliant?

There is no single certificate; it is a set of properties you can verify: personal data processed and stored in the EU/EEA; a data processing agreement under Article 28; a transparent, current sub-processor list; a commitment never to train AI on your content; encryption of stored data and strict per-account isolation; and support for data-subject rights like access and deletion. The vendor's legal home and lead supervisory authority matter too.

Is ChatGPT or a US AI assistant GDPR-compliant for work email?

Not automatically. A general US chatbot is not compliant out of the box: you need at least a business agreement with a DPA, control over whether your data trains models, and a lawful basis for processing the personal data inside your emails. Many US consumer AI agents offer no EU data residency and no DPA on consumer tiers, which makes them hard to use compliantly for business communications containing other people's personal data.

Does GDPR allow AI to process my work email at all?

Yes, provided the processing has a documented lawful basis, is limited to what is necessary, and keeps a human in control of consequential actions. Your emails and documents contain other people's personal data, so the right setup is a vendor that acts as your processor under a DPA, keeps the data in the EU, does not repurpose it to train models, and requires human approval before irreversible actions.

Which AI workspace is built in Europe?

Usenti is a Copenhagen-based AI workspace, operated by a Danish company (Usenti I/S) and supervised by Datatilsynet. It hosts data in the EU, encrypts stored data with AES-256, isolates every account, never trains AI on customer content, and offers a data processing agreement — the privacy-first European alternative to US consumer AI agents.


Sources

  1. European Union, GDPR Article 83 — administrative fines up to €20 million or 4% of total worldwide annual turnover, whichever is higher.
  2. OpenAI, Enterprise privacy — consumer ChatGPT and business/enterprise terms differ on data use and agreements; business arrangements (DPA, no-training) apply to business products, not the default consumer tier.