The European Alternative to Muse, Instinct, Grok Bot and dots

Why Europe needs its own AI assistant

A personal AI agent is the most privacy-sensitive software most people will ever install. To be useful, it has to read your inbox, your calendar, your meeting transcripts and often your messages. It sees your clients' names, your health appointments, your contracts and your family's travel plans. Under the GDPR, that data comes with obligations: a lawful basis for every use, a named controller, a processing agreement with every vendor that touches it, a defined place where it lives, and a right for every person in it, including the people who email you, to see it and have it deleted.

The four agents dominating the headlines this autumn were built for a different legal environment. That isn't a judgement on the engineering, which is impressive in all four. It's a question of what happens to a European's inbox once an American agent has read it, and whether anyone has written down the answer.

The four agents, and where each one leaves a European user

Meta's Muse: not launched in the EU

Muse is Meta's personal agent, launched in the United States on September 8, 2026 and in Canada on September 18, running on Meta's own Muse Spark model. It sends emails, books travel, fills in forms and makes purchases, with a separate approval agent watching sensitive actions, and it is available in iOS, Android, Mac and web apps and through WhatsApp. It reached roughly 2.8 million installs in its first twelve days.

For Europeans, the short version is that it isn't available. An independent tracker looking at Muse and the EU found no launch date, no data processing agreement and no EU data residency. Interactions train Meta's models by default, with an opt-out, and Meta's own documentation states that its isolation "does not prevent Meta from accessing data when necessary." A fully private mode has been promised but not shipped. For a business in Europe, a tool with no DPA cannot lawfully be pointed at an inbox full of other people's personal data.

Instinct: trains by default, says nothing about where data lives

Instinct, from Spear Street Technology, is the agent that texts you first. It lives in iMessage, WhatsApp, SMS and phone calls, has its own phone number and cloud computer, and will check you in for flights, cancel subscriptions and negotiate bills. It has raised about $1.25 billion, most recently a $1 billion Series C at a $10 billion valuation.

Its privacy terms are where a European reader should slow down. User data trains Instinct's models by default; the opt-out applies only to future data. The terms grant Instinct a "perpetual and irrevocable" licence to user materials. The privacy policy names no storage location, no retention period, no model providers and no GDPR basis or transfer mechanism. TechCrunch and others have reported inbox data kept after users disconnected, and an email sent without approval. The Next Web has also raised questions about how its outbound phone calls fit the EU AI Act's disclosure duties.

Grok Bot: always-on, but no documented EU commitments

Grok Bot is xAI's "always-on AI teammate", launched in beta in August 2026 and distributed together with Cursor. Each bot gets a persistent cloud computer with a browser, terminal and file system, can run scheduled routines, and can learn a task from a screen recording. It ships in Mac, Windows and iPhone apps and comes bundled with SuperGrok Heavy or Cursor Ultra rather than as a standalone plan.

On data, it follows Cursor's account and privacy settings, cloud storage is mandatory, and the retention period is not documented. There is no way to inspect or delete its memory, and credentials remain on the machine after a bot is deleted. Its own docs warn not to treat separate bots as a security boundary. We could not find documented EU availability, a DPA, or a stated data location for Grok Bot specifically. With no spending cap on on-demand usage, it is also the hardest of the four to budget for.

OpenAI's dots: the most governed, but gated in Europe

dots, launched by OpenAI on September 29, 2026, runs inside ChatGPT on web, desktop and mobile, and in Slack and Teams. Each dot has its own cloud computer, works toward goals in the background, reaches more than 4,000 apps through connectors, and checks risky actions with an auto-review step. OpenAI does not train on Business, Enterprise or Edu data by default.

Of the four, dots has the most serious governance story, and it deserves that credit. The European catch is access. Your first dot comes with ChatGPT Pro ($100 to $500 a month) or a Business Premium seat (about $100 per user per month billed annually). ChatGPT Pro is not offered in the EEA, Switzerland or the UK, so a European individual cannot get dots on a personal plan at all. The route in is a Business Premium seat, and even then the data still sits with a US provider under its standard business terms.

What makes Usenti the European option

Usenti is an AI workspace for your email, calendar, meetings and to-dos, with an assistant, Donna, who reads all of it and does the work: drafts the reply, books or moves the meeting, writes up the call, turns the ask in last week's thread into a to-do. Where it differs from the four agents above is not one privacy feature. It's that every decision about your data was made with the GDPR as the starting point.

1. A European company, a European regulator

Usenti is incorporated under Danish jurisdiction. Its lead supervisory authority is Datatilsynet, the Danish Data Protection Agency, and users elsewhere in the EU or EEA can go to their own authority. That means a European regulator can actually reach the company holding your data, and Europe isn't an afterthought market for a launch that happened somewhere else first.

2. Your data stays in Europe by default

Usenti's database is hosted in the EU region. Backend hosting includes Hetzner, an EU-based provider. Product analytics, used only if you consent to analytics cookies, run on PostHog's EU Cloud, and typed input and email content are excluded from session recording. The meeting notetaker records in the EU (Frankfurt) region by default, with the US available only if you pick it per account. Where a vendor operates in the US, such as Google's AI services or Vercel for the website, the transfer rests on the European Commission's Standard Contractual Clauses inside that vendor's data processing agreement. That is the honest picture: not "nothing ever leaves Europe," but every transfer named, contracted and lawful.

3. Your content never trains a model

Mail, messages and transcripts are sent to an AI model only to produce a result for you. They are not used to train, tune or evaluate models, by Usenti or by its providers, who process them under a data processing agreement. Meeting content goes only to Google Cloud Vertex AI in Usenti's own project, never to a consumer AI API. Compare that with Muse and Instinct, where training is the default and privacy is the opt-out.

4. A processing agreement for every vendor, published

Usenti maintains a data processing agreement with every sub-processor, as GDPR Article 28 requires, and publishes the list: Supabase, Render and Hetzner, Vercel, Stripe, Google, Microsoft, Recall.ai, Deepgram and PostHog, each with what it does and where. Every new vendor is assessed against five questions before adoption: independent security assurance, a DPA, where data is stored and processed, deletion on request, and breach notification. Before you connect an inbox you can read exactly who will touch it, which you can't do with Muse or Instinct today.

5. Rights for the people in your inbox, not just for you

An AI assistant processes the data of everyone who emails you, not only yours. Usenti treats them as data subjects with rights. Anyone, user or not, can file an access or deletion request through a public form. Requests land in a worked queue that acknowledges the requester, escalates as the one-month deadline under Article 12(3) approaches, and can't be closed without recording what was done. The meeting notetaker announces itself in the meeting chat, says whose notetaker it is and where to object, deletes the recording once it's transcribed, and rests on a written legitimate-interests assessment.

6. Nothing irreversible without your click

Donna does the work rather than describing it, but with a rule in place. Anything reversible, such as creating a to-do or moving it, runs at once and comes back with an Undo. Anything that can't be taken back, such as sending an email, cancelling a meeting that has attendees, or any delete, is prepared on a card with the real recipient and the actual words. It runs only when you press the button. Email bodies and meeting transcripts are treated as untrusted data, never as instructions, so the worst a malicious email can do is put a card in front of you that you decline. That is the opposite of an agent that has already sent an email by the time you read about it.

7. Independent, and in your language

Usenti isn't a feature of an advertising business, a social network or a model lab. It sells a subscription and nothing else, so your inbox isn't an input to anything you didn't ask for. And because it was built in Europe, Donna answers in the language you address her in, written or spoken. Write to her in Danish and the reply comes back in Danish; ask in German by voice and she answers in a German voice. Meeting transcription detects the language spoken, including Danish, Swedish, Norwegian and Finnish, and writes the notes in it.

Usenti vs Muse vs Instinct vs Grok Bot vs dots, at a glance

Every row reflects what each company has published, or what independent coverage has reported, as of early October 2026. These products are weeks old and changing fast, so a dash often means "not documented" rather than "impossible."

✓ = confirmed / built in  ·  ~ = partial, conditional or approached differently  ·  – = not offered or not publicly documented. Competitor rows reflect vendor pages and the independent coverage cited at the end of this article; Usenti rows reflect Usenti's published Privacy Policy, Compliance page and product.

What Usenti does, day to day

Being GDPR-compliant only matters if the product is worth using. Here's what you get.

• A smart, unified inbox across Gmail, Outlook / Microsoft 365, Zoho Mail, AWS WorkMail and any IMAP inbox, plus Slack and Microsoft Teams, sorted in real time into categories like To Respond, Awaiting Reply and Important.
• Draft replies in your tone, adapted to how you write to that specific person, in the language the message arrived in.
• Donna, your assistant: a morning brief over your calendar, inbox and to-dos (which she'll read aloud), and a chat on every page where you can ask "what did Marie send last week?" or "move my 3pm to Thursday" and have it done.
• A meeting notetaker for Google Meet, Microsoft Teams, Zoom and Slack huddles, EU-region by default, with notes, decisions and action items in the language spoken.
• One calendar across Google, Outlook and Apple, with booking links that never double-book you.
• To-dos that find themselves: asks in your mail and meetings become suggested to-dos that wait for your approval rather than appearing on your board unannounced.

Who should choose which

If you live in the US, want an agent to buy things and book travel, and are comfortable with Meta's data practices, Muse is the most polished consumer agent today. If you're a heavy Cursor user who wants autonomous routines on a cloud computer, Grok Bot is built for you. If your company already runs on ChatGPT Business and can justify around $100 a seat, dots brings real governance.

If you're in Europe, or you work with people who are, and you want an AI assistant that reads your email, runs your calendar and writes up your meetings without sending your inbox into a training set or a legal grey zone, Usenti is the one built for you.

Running a team rather than just your own inbox? Read Usenti: the business version of Muse, Instinct, Grok Bot and dots. For the wider field of email-first tools, see the best AI email assistants in 2026, ranked.

Frequently asked questions