How we protect your data, in plain words.
How local AI and filtered cloud requests fit together: read our approach.
The AI models, the document archive and the memory run on the computer in your office. Data can stay on your premises: nothing leaves without a path you decide. The cloud is optional. Remote support connects to the computer only at your request and with your consent.
The owner decides who may see and act on the computer. Every application has an emergency encryption button for administrators: it locks the data and destroys the keys kept next to the computer. Data comes back only with the backup keys you stored beforehand.
The computer keeps regular copies of the archive, so you can start again after a fault.
We connect remotely only at your request and with your permission. During the connection, encrypted personal data stays disconnected and is not accessible to support. Every request becomes a ticket.
If you find a security vulnerability in our computers, software or website, write to us. Describe what you found, where it is and how to reproduce it, and leave a contact so we can reach you.
We confirm receipt within 4 working hours at most. We agree on publication with you once a fix is ready (coordinated disclosure), and we take no legal action against anyone who researches and reports in good faith.
Where the Cyber Resilience Act requires it, we report actively exploited vulnerabilities and serious incidents to ENISA, within our obligations as a manufacturer.
A useful report includes: the affected product, version, page or address; the steps to reproduce; the expected and the observed behaviour; the possible impact; screenshots or a proof of concept, if available.
Good faith means: you access only the data needed to show the problem, you do not change or delete data, you do not degrade the service (no denial-of-service or spam tests), you do not use social engineering or physical attacks, and you keep the details confidential until we agree on publication.
Machine-readable contact (RFC 9116): /.well-known/security.txt
External models only through a filter: personal data and trade secrets are removed while keeping the meaning of the request; when needed, you see the filtered request and confirm it before it goes to the cloud.
The filter helps reduce risks; it does not promise 100% anonymisation.
