Using a personal AI assistant in a GDPR-compliant way starts with one question: where does the data go when your team writes to it? Monday morning. Lisa, the operations lead at a 40-person agency, finds out that three people on her team have been pasting client briefs into a free AI chat tool for weeks - no DPA, no idea whether the data trains a model, no record of what went in. The clients are in Germany, the data is personal, and the next audit is in four weeks. This guide shows, step by step, how to hand work to an AI assistant without that moment - and how Superkind’s ready-made AI team for compliant work does it from day one.
Key takeaways
Get a ready-made AI team for compliant work - Superkind comes with an AI team (Inbox Guard, Compliance Tracker, Data Request Handler) that handles email, DPA tracking and data subject requests automatically. Start for free with unlimited users, or build your own.
Keep control, keep data in Europe - Superkind is hosted in Europe, GDPR compliant, never trains on your data. You decide which steps need your approval before anything goes out.
Six steps to compliance - check the DPA, verify EU hosting, opt out of model training, set approval rules for sensitive tasks, build an audit trail, train your team on AI literacy (Art. 4 EU AI Act).
Shadow AI is the real risk - teams already use AI tools without telling IT. The fix is not to ban AI but to give them a compliant one that works in the channels they already use.
Your AI team for GDPR-compliant work
What does GDPR-compliant AI mean?
A GDPR-compliant AI assistant is one that processes personal data under a signed data processing agreement (DPA/AVV) per Art. 28 GDPR, hosts and processes data inside the EU, does not train its models on your inputs, and gives you a way to fulfil access, deletion and objection requests at any time. The compliance sits in the contract and the infrastructure, not in the AI itself. One example is Superkind: an AI employee that works in Outlook, Teams and Slack, hosted in Europe, with a DPA through the Trust Center and approval where you want it.
The difference matters: a team lead who pastes customer names into a free chat tool has no DPA, no deletion path and no audit trail. A team that writes to Superkind in Teams has all three, because the infrastructure was built for it.
How it works with Superkind: your AI team for GDPR-compliant work
You write to Superkind by email, in Teams, in Slack or in the web app, like you would write to a colleague, and Superkind does the task in your programs - over 1,700+ integrations, including Outlook, HubSpot, Notion and Zendesk. It works with your tools instead of replacing them, hosted in Europe and GDPR compliant.
With Superkind you don't start from zero: it comes with ready-made AI employees for compliant day-to-day work. They run automatically in the background, and you decide where they ask you first. Use them right away, adapt them, or build your own.
Inbox Guard
AI employee for email triage
Support emails pile up and every one carries a name, a contract number, sometimes health data. The Inbox Guard reads incoming mail, sorts it by topic and urgency, drafts a reply from your templates in Outlook and routes the rest to the right person in Teams. Personal data stays in Europe, and anything with sensitive content waits for your OK before it goes out.
Compliance Tracker
AI employee for GDPR compliance
DPAs expire, deletion deadlines pass, and nobody notices until the audit. Every Monday the Compliance Tracker checks your contracts in SharePoint, flags DPAs that expire within 90 days, lists tools without a signed AVV and reminds you which deletion requests are overdue. No approval needed - it just reports.
Data Request Handler
AI employee for data subject requests
A customer asks what data you hold on them. You have 30 days under Art. 15 GDPR. The Data Request Handler collects the data from HubSpot, Notion and Outlook, compiles it into one export, drafts the reply letter and sends it to you for review. You approve, it goes out.
Every employee is ready to use right away. Or build your own on the platform.
Safe by default
Personal data in emails, CRM records and support tickets stays in Europe. Superkind never uses it to train a model.
- Hosted and processed in Europe
- GDPR compliant
- Aligned with the EU AI Act
- No training on your data
- DPA through the Trust Center
What you need
Checklist before you start
- A list of AI tools your team already uses (including free ones - this is your shadow AI audit)
- Your data processing register (Verarbeitungsverzeichnis, Art. 30 GDPR) or the person who maintains it
- Access to your vendor contracts and DPAs
- An agreed approver for tasks that touch sensitive personal data (HR records, health data, financial data)
- A free Superkind account at app.superkind.ai - the DPA is included in every plan
Step by step: making your AI assistant GDPR-compliant
Step 1: Check the DPA (Art. 28 GDPR)
Before your team writes the first message to any AI assistant, check whether a data processing agreement (DPA/AVV) is in place. Art. 28 GDPR requires a written contract with every processor before processing begins1. The DPA must name the purpose, the categories of personal data, the duration, the subprocessors and your audit rights.
- What to look for - a signed DPA that lists the specific processing purposes, not just “to provide the service”. It must name subprocessors and give you audit rights.
- Red flag - no DPA available, or a DPA that only exists on enterprise plans. If the free or standard plan has no DPA, any personal data your team puts in is a compliance gap4.
- Superkind - DPA available through the Trust Center at trust.superkind.ai, included in every plan including the free one.
Step 2: Verify EU hosting and processing location
Where does the data go when your team writes to the AI? If the answer is “US servers”, you need Standard Contractual Clauses (SCCs) under Art. 46 GDPR or a valid adequacy decision. The simplest path: choose a provider that hosts and processes inside the EU.
- Check - ask the vendor where inference (the actual AI processing) happens, not just where data is stored. Some vendors store data in the EU but send it to US servers for processing.
- Superkind - hosted and processed in Europe. AI models from OpenAI, Anthropic, Mistral and Google, all processed in the EU. No SCCs needed.
Step 3: Opt out of model training
Many AI providers train their models on user inputs by default. Under GDPR, using customer data to train a third-party model requires a separate lawful basis - and your customers almost certainly did not consent to it. Check the vendor’s terms for a training opt-out, and confirm it is active on your account.
- What to look for - an explicit, written confirmation that your data is not used for model training. A toggle in the settings is not enough if the default is on and nobody switched it off7.
- Superkind - does not train on your data. This is not a toggle you need to find - it is the default. Confirmed in the DPA.
Step 4: Set approval rules for sensitive tasks
Not every task needs a human check. But when the AI assistant drafts a reply that contains a customer’s financial data, processes an HR record or sends a document externally, someone on the team should see it first. Set rules for which types of tasks wait for approval.
Rule: every outgoing email that contains personal data waits for my approval. Internal summaries run on their own.
Superkind applies this rule across Outlook, Teams and Slack - the same approval logic in every channel. Routine work (sorting mail, updating a CRM field, posting a summary) runs without asking. Sensitive work waits.
Step 5: Build an audit trail
Art. 5(2) GDPR requires you to demonstrate compliance - not just be compliant1. That means you need a record of what the AI processed, which data it accessed, what it produced and who approved it. If a supervisory authority asks “what happened to this customer’s data?”, you need an answer.
- Minimum - every task the AI handles is logged with a timestamp, the data categories involved and the outcome (sent, rejected, pending).
- Superkind - every action is in the audit log: which systems it accessed, what data it read, what it produced, who approved it. Available through the web app and the Trust Center.
Step 6: Train your team - AI literacy (Art. 4 EU AI Act)
Since 2 February 2025, Art. 4 of the EU AI Act (Regulation (EU) 2024/1689) requires every company that uses AI to take measures promoting AI literacy among its staff2. This is not a certificate or a one-off course. It means: your team knows what the AI can and cannot do, what data they may and may not give it, and where the approval points are. Document who received training and when.
- Practical step - a 30-minute session with your team: here is our AI assistant, here is what it does, here is what needs approval, here is what you must not paste into a free tool. Write it down.
- Tip - Superkind’s approval rules make the training concrete: “If you see an approval card, check the data before you hit Approve. If it looks wrong, hit Reject and tell Superkind what to fix.”
Set it up as a routine
Once the six steps are in place, compliance stops being a project and becomes a routine. Tell Superkind what should happen on a schedule, and it keeps running.
Every Monday at 8:00: check all vendor DPAs in SharePoint, flag any that expire within 90 days, and list tools without a signed AVV. Post the report in the Compliance channel.
- Weekly - DPA status, overdue deletion requests, tools without an AVV.
- On every incoming data subject request - collect the data from HubSpot, Notion and Outlook, compile the export, draft the reply, wait for approval.
- On every outgoing email with personal data - check the approval rule, wait if required, log the action.
When the process grows - for example, different rules per department, a custom compliance dashboard or an integration with your DMS - you do not start over. Build custom solutions on the Superkind platform, on top of what it already knows. The 1,700+ integrations keep everything connected to SharePoint, HubSpot, Outlook and Notion without new middleware.
Data protection and EU hosting
Every task your AI assistant handles touches data that matters: customer names, contract numbers, email addresses, sometimes financial or health data. The question is not whether to protect it, but whether your setup already does. With Superkind, approval where you want it means sensitive tasks wait for a person. The sovereign AI explainer covers what sovereignty means in practice, and AI tools built in Europe lists the providers that host and process inside the EU.
Other ways to use AI assistants compliantly
| Way | Effort | Risk | Best for |
|---|---|---|---|
| Superkind | Low: DPA included, EU hosting, approval built in | Low: audit trail, no training, 1,700+ integrations | Teams handing real work to AI in Outlook, Teams or Slack |
| ChatGPT Enterprise | Medium: sign DPA, configure EU region, verify settings | Medium: check training default, SCCs may apply | Large teams already on OpenAI with IT support |
| Self-hosted LLM | High: set up infrastructure, maintain updates | Low if done right, but needs dedicated staff | Regulated industries with in-house AI teams |
| No AI | None | Shadow AI risk: teams use free tools anyway | No one - banning AI does not stop it |
Common mistakes
- Assuming “EU server” means compliant - storage in the EU is not enough if inference happens in the US. Check where the actual AI processing runs, not just the data centre label.
- Ignoring the free plan - if three people on your team use a free AI chat tool without a DPA, you have three uncontrolled processing activities. Shadow AI is the most common GDPR gap in 20268.
- Skipping the training opt-out check - some providers train on inputs by default and bury the opt-out in settings. If nobody switched it off, your customer data may already be in the training set.
- No approval for outgoing messages - an AI that sends a reply containing personal data without anyone checking it is a data breach waiting to happen. Set approval rules for external communication.
- No audit trail - “We use AI responsibly” is not enough. Art. 5(2) GDPR requires you to prove compliance. Without a log, you cannot.
6 tips for GDPR-compliant AI assistants
1. Start with the shadow AI audit
Before you set up anything new, find out which AI tools your team already uses. Ask directly, check browser extensions and app subscriptions. Every tool without a DPA is a gap you can close today.
2. Use the channels your team already works in
If your team lives in Outlook and Teams, give them an AI assistant that lives there too - like Superkind. A separate login is a barrier, and barriers push people back to the free tool they were using before.
3. Make approval visible, not hidden
Approval should show up where the person works: in Teams, in Slack, in Outlook. Not in a separate dashboard nobody opens. Superkind sends approval cards in the channel where the approver is.
4. Document the AI literacy training
Art. 4 EU AI Act does not require a certificate, but it does require documented effort5. Write down: date, attendees, topics covered (what the AI does, what data it may process, where the approval points are). Keep the record.
5. Review the DPA when the vendor changes terms
AI vendors update terms frequently. Set a reminder to re-check the DPA, the subprocessor list and the training opt-out every quarter. Superkind’s Compliance Tracker can automate this.
6. Test with real tasks before you roll out
Give the AI assistant three real tasks from last week - a support email, a data request, an internal summary. Check where the data goes, what the audit trail shows, and whether the approval rule fires. Fix anything before the team starts.
Conclusion
GDPR-compliant AI is not about choosing between productivity and protection. It is six concrete steps: sign the DPA, verify EU hosting, opt out of training, set approval rules, build the audit trail, train your team. The fastest way to see it working: sign up for free at app.superkind.ai, write Superkind your first task in Teams or Outlook, and check the audit log. If you want to go deeper, the sovereign AI explainer covers what European sovereignty means for your data.
Frequently Asked Questions
Three things: a signed data processing agreement (DPA/AVV) under Art. 28 GDPR with the provider, processing and hosting inside the EU so no transfer mechanism is needed, and a confirmed opt-out from model training on your data. On top of that, you need a lawful basis under Art. 6 for every type of personal data the assistant processes, and a way to fulfil deletion and access requests.
Yes, if the tool processes personal data on your behalf. Art. 28 GDPR requires a written agreement before processing begins. Many consumer-tier AI plans do not offer a DPA at all, which means any personal data your team puts in creates a compliance gap. Check the vendor's trust or legal page for a DPA before you start.
Superkind is hosted and processed in Europe, GDPR compliant and aligned with the EU AI Act (Regulation (EU) 2024/1689). It does not train on your data, and the DPA is available through the Trust Center at trust.superkind.ai. AI models from OpenAI, Anthropic, Mistral and Google are processed in Europe. You decide which steps need approval before Superkind sends, books or changes something.
It depends on the plan. Consumer and free tiers typically have no DPA and may train on your inputs. Enterprise plans from OpenAI and Microsoft offer DPAs and EU data residency options, but you need to check the exact configuration - for example, whether data stays in the EU region, whether training is off by default and whether subprocessors are listed. Always verify the vendor's current terms before you start.
Since 2 February 2025, Art. 4 of the EU AI Act (Regulation (EU) 2024/1689) requires every company that uses AI to take measures promoting AI literacy among staff. This is not a certification but an ongoing duty: document what training you provide, who received it and when. Fines are enforced by national authorities from August 2026.
You start for free at app.superkind.ai with up to 150 hours of work free, no credit card and no sales call. The Business plan starts at EUR 50 per month for 50 hours of work for the whole company with unlimited users, never a price per seat. The DPA, EU hosting and the training opt-out are included in every plan, including the free one. As of November 2026.
Related Articles
- Sovereign AI Explained: What It Means for European Companies
- AI Tools Built in Europe: The Complete Guide
- Is Superhuman GDPR-Compliant?
Sources
- EUR-Lex - Regulation (EU) 2016/679 (GDPR), Art. 28: Processor
- EUR-Lex - Regulation (EU) 2024/1689 (EU AI Act), Art. 4: AI Literacy
- IHK Frankfurt - KI und Datenschutz: AVV, DSFA und Rechtsgrundlagen
- Compound Law - DSGVO-Checkliste für KI-Tools in deutschen Unternehmen
- Traverse Smith - The EU AI Act's AI Literacy Requirement: Key Considerations
- European Commission - AI Act Implementation Timeline
- Advisori - GDPR-Compliant AI: On-Premise and EU-Sovereign Models for Enterprises (2026)
- Computerwoche - DSGVO-konforme KI-Workflows in der Praxis (2026)
Write Superkind your first GDPR-compliant task.
Start for free in Outlook, Teams or Slack. Or let us show you in 30 minutes how it runs in your systems.

