Running support/MivoChat field guide

Choose a GDPR-conscious AI support chatbot

A SaaS team evaluating AI support should check hosting, data processing, retention, and model handling against its own obligations.

By MivoChat product team · Choose a GDPR-conscious AI support chatbot · Published and updated October 12, 2026

Quick answer

Review where data is hosted, the DPA, subprocessors, retention, access controls, and whether customer content is used for training. MivoChat is hosted in Frankfurt, the DPA is available at /dpa, subprocessors are listed at /subprocessors, and customer data is not used to train models.

Installation steps

    Ask for the relevant documents

    Review the DPA, subprocessor list, retention behavior, and security information with whoever owns privacy at your company. EU hosting can help with data location, but does not by itself make a deployment compliant. Map the data path from the visitor form to the conversation record, connected support channel, and model provider; note what fields are copied at each handoff and which roles can view them. Compare that inventory with your own notice and internal records before turning the widget on for a public audience.

    Limit sensitive content

    Do not invite users to paste passwords, API secrets, or private personal data into support chat. Document what the team should do if a visitor shares sensitive information and limit access to conversation records. Make the warning visible beside the chat entry point, decide who can redact or delete a message, and define how staff should route a request that contains credentials. Review connected Telegram or Slack channel membership too, because chat data can be visible to everyone who can read that destination.

    Before you change the setup

    a privacy owner, data-flow map, intended chat fields, retention policy, and the company's legal basis and notices

    the DPA at /dpa, subprocessors at /subprocessors, and the product's current hosting and retention information

    Frankfurt hosting and a DPA are useful facts but do not by themselves determine whether your specific processing is compliant

    Test the support boundary

    • Where is chat data hosted and which subprocessors handle it?
    • How long are conversations retained and who can access them?
    • Does the assistant use our conversations to train a model?

    Frequently asked questions

    A privacy review assumes EU hosting resolves every transfer

    Map the full processing flow, including model and support subprocessors, then assess it with your privacy owner.

    Visitors paste credentials into a chat

    Add a visible instruction not to share secrets and define a process for redacting or handling sensitive messages.

    The privacy notice does not mention chat

    Update your notice and internal record of processing to describe the actual purposes, recipients, and retention for your deployment.

    A subprocessor review relies on an old export

    Use the current /subprocessors page and DPA when approving the service, and repeat the check when your vendor review cycle requires it.

    Keep exploring