GDPR-friendly AI: keep prompts and personal data on your own servers.
Under the GDPR, every prompt that contains personal data is a processing activity. Sending it to a hosted model provider means a processor, a transfer assessment and a lot of trust. Running open models on your own servers keeps that data under your control, in the location you choose.
The problem with hosted AI and personal data
When an employee pastes a customer email, a CV or a medical note into a hosted chatbot, that personal data is processed by the AI provider. For your data protection officer that raises questions:
- Who is the processor, and is there a data processing agreement that covers this use?
- Where is the data processed, and does it leave the EU?
- How long is it retained, and is it used to train or improve models?
- Can you honor access and deletion requests for data inside prompts?
Enterprise plans answer some of these questions contractually. Self-hosting answers them technically: the data never goes there.
How Loadcore keeps data in place
Loadcore splits the platform into two paths:
| Inference path | Control path | |
|---|---|---|
| Contents | Prompts, completions, streaming output | Heartbeats, download progress, instance status, key checks |
| Runs on | Your GPU servers | The managed Loadcore control plane |
| Personal data in prompts | Stays on your servers | Never received |
Model weights download from Hugging Face straight to your server. The control plane knows which servers and models exist and who may use them, nothing more.
Data minimization by design
- Per-department access: members only reach the models their department is granted, which supports purpose limitation.
- Personal API keys: every request is tied to a person, keys are shown once, stored hashed and can be revoked at any time.
- No prompt storage in the platform: Loadcore has no copy of your prompts that could leak, be subpoenaed or be used for training.
- Analytics without cookies: the Loadcore website and console use EU-hosted analytics with text masking and no cookies.
Choose your location
Because inference runs where the agent runs, you decide the jurisdiction. Use servers in your own data center, at a European colocation provider or in an EU region of the cloud provider you already use.
For your DPO
Our privacy page lists exactly what Loadcore stores today. For questions about the data Loadcore processes, email privacy@loadcore.io.
Related
Learn how on-premise LLM hosting works with Loadcore, or see what private AI looks like for healthcare and law firms.
Frequently asked questions
Does using Loadcore make us GDPR compliant?
No tool makes you compliant by itself. Compliance depends on your purposes, legal bases, retention and security measures. What Loadcore changes is the data flow: prompts and completions are processed on your servers and never sent to Loadcore or to a model provider, which removes a large part of the assessment.
What personal data does Loadcore itself process?
To run the service, the control plane stores account data (name, email and sign-in method), your organization structure, server details reported by the agent (hostname, port, public IP, operating system and GPU) and API key metadata. Keys are stored hashed. It never receives prompts, completions or model weights.
Can we choose where inference runs?
Yes. Inference runs on whatever server you start the agent on: your own data center, a colocation facility or a GPU server in a cloud region you choose, for example inside the EU.
What about the EU AI Act?
Most obligations in the EU AI Act depend on what you use AI for, not where it runs. Self-hosting does give you full control over which model version is in use, who can access it and how it is configured, which makes documenting your use easier.
Run your first model today.
Register a GPU server, pull an open model and hand your team an API key.