Where does computation happen?
Local.
Prioritise processing on the device to reduce the need to send personal information to a remote service.
Intelligence
at your service.
In your hands.
EVA-I is a French company developing personal, local AI.
Our goal: useful intelligence, with your data under your control.
The company
AI that helps you.
Context that belongs to you.
EVA-I is developing artificial intelligence designed to work close to the person using it: on their device, with the context they choose to share.
Our project is to make AI useful in everyday life: understanding a request, finding information in personal documents and taking preferences into account. Local processing is central to this approach.
Development is ongoing. Available features, limitations and conditions of use need to be documented before any availability announcement.
About EVA-IThree design principles
Local. Personal.
Sovereign.
Where computation happens, whom AI adapts to and who decides how data is used: three distinct questions, connected in our botanical symbol.
01 / The locationLocal
02 / The contextPersonal
03 / The decisionSovereign
Three connected principles.
One approach.
Whom does AI adapt to?
Personal.
Take into account the documents, preferences and conversations that the person chooses to make available.
Who stays in control?
Sovereign.
Give people control over their context and the external services they choose to use.
Europe and data protection
Developing in Europe.
Designing around
people’s rights.
Data protection is part of the technical problem to solve. It should guide architecture decisions and how those decisions are explained.
About EVA-IA French company.
EVA-I takes a European perspective: developing useful tools while respecting each person’s control over their data.
Protection by design.
The GDPR provides for data minimisation and data protection by design. Local processing can contribute to this, but does not establish a service’s compliance on its own.
Explain the dependencies.
Data sovereignty also means understanding which components are used and what information leaves the device when an external service is involved.
A European approachControl over personal dataLocal · Personal · Sovereign
Method and foundations
Clear principles.
Verifiable evidence.
Local inference, document retrieval and model adaptation build on documented methods. Our technical documentation is being prepared.
Read our methodEssential questions
What our approach
means.
Can local AI make mistakes?
Yes. Running a model on a device does not guarantee accurate answers. Important information needs to be checked, and the limitations of each feature need to be evaluated.
Does local mean the internet is never used?
No. Downloads, updates or the use of an external service may require a connection. These data flows need to be described in the documentation for each feature.
Where can I find EVA-I’s scientific results?
The Research page is being prepared. External references on the Method page are distinguished from EVA-I’s own results; they do not validate our developments.
Talk to EVA-I
A question
about our approach?
Let’s talk.
To understand our approach, ask a technical question or propose a conversation, contact the EVA-I team.
Contact EVA-ILocal.
Personal.
Sovereign.