Support and salesSupport / sales
Unlimited Basic and ProFrom €14.95/month · Unlimited models subject to fair use

Security

Privacy by design. Control over your data.

Security built into every request and deployment.

QDivZero is a platform developed and operated by Valendra Tech SL, a Spanish company. We apply security controls to protect your data during inference, from encrypted communications to deployment isolation and our zero logs policy.

Your prompts and responses are not stored by default or used to train models. The Responses API is the exception, retaining the context needed to continue conversations.

Privacy does not depend on the model you use or the region where you run it. We maintain our inference content protection policy across all available regions.

01 /Security controls

Zero logs

Zero logs by default.

We do not log inference content by default. Responses retains context for conversations; you can delete it or disable storage and use Chat Completions.

Deployment isolation

Isolated environments.

Each deployment keeps its execution environment, resources, and configuration separate from other deployments.

Dedicated infrastructure

Dedicated GPUs.

Your model uses exclusive GPU resources while the deployment is active, without sharing inference capacity with other deployments.

Encryption in transit

Encrypted data.

HTTPS/TLS protects prompts, files, and responses between your application and QDivZero, and between QDivZero and the inference infrastructure.

No training

No model training.

We do not use your prompts or responses to train models or send them to their creators. This policy applies in every region.

Data control

Data under control.

Responses retains context to continue the conversation. We do not use it for other purposes, including model training.

Region control

Developed in Europe. Deploy where you need.

QDivZero is developed and operated by a Spanish company and is governed by the European data protection framework.

Every application has different needs. You can therefore choose where to run your models from the available regions in Europe, North America, Asia, Oceania, and Africa.

Our multi-provider network automatically finds compatible infrastructure in your selected region, without requiring you to manage each provider directly.

Unlimited models in Unlimited Basic and Unlimited Pro run exclusively in Europe; public models and subscription models that use AI Credits run in Europe and the United States. For your own deployments, choose from the available locations while keeping our zero logs by default policy.

Explore providers and regions

02 /European framework

GDPR

Data protection under the European framework.

A platform developed and operated by a Spanish company under the European data protection framework.

Region control

Decide where your models run.

Choose from the available regions based on your application needs and your organization’s location requirements.

Governance

More transparency over your AI infrastructure.

Identify the models, providers, and resources in your deployments to support your oversight and governance processes.

Protection before inference

With LLM Firewall, you can analyze requests to detect prompt injection, jailbreaks and sensitive data before inference. Combine built-in rules and custom policies to control which requests reach your models.

Provider transparency

Public models are served through Azure, GreenPT, OVH and Cerebras. Private deployments use on-demand GPUs from Runpod and Vast.ai. The provider executing the request generates the result.

Transparency

Know what happens to your data.

From your application to the model, with visibility at every step.

Each request passes through several components before the model generates a response. At QDivZero, we want you to understand the role of each component and how we protect information along the way.

Your application connects to QDivZero using HTTPS/TLS. The platform directs the request to the corresponding endpoint or deployment, and the inference infrastructure runs the model.

Responses return over encrypted connections. You can also add LLM Firewall to analyze requests before they reach the model.

View the inference architecture

03 /Your data path

Two encrypted connections, in both directions
  1. 01

    Your application

    Prepares the request

    Sends the model, instructions and data needed for the task.

  2. 02

    QDivZero

    Connects to the destination

    Directs the call to your selected destination. If you enable Paranoia Mode, detected sensitive data is masked before sending.

  3. 03

    Inference infrastructure

    Runs the model

    Processes the request and generates the result. Our provider does not store inference data; with Paranoia Mode, it receives less sensitive information.

Requests travel encrypted using HTTPS/TLS between your application and QDivZero, and between QDivZero and the GPU or public inference provider. Responses return encrypted over those same two connections. Encryption protects content in transit.

AN OPTIONAL LAYER ALONGSIDE ZERO LOGS

Paranoia Mode. Protection you choose to enable.

We apply zero logs at QDivZero by default, and our provider does not store inference data. As an extra option, you can enable Paranoia Mode so we mask detected passwords and sensitive personal data on our servers before sending the request.

Even without storing data, the provider needs to process it to respond. Sending less sensitive information reduces exposure during processing. If you enable this option, it complements zero logs and HTTPS/TLS encryption, which remain in place with Paranoia Mode disabled.

Explore Paranoia Mode

QDivZero receives the originalThe provider receives the protected version

04 / FAQ

Frequently asked questions about security and privacy

Answers to common questions about zero logs, data privacy, encryption in transit, region selection, and running AI models privately.

Do you send my data to model providers, including outside Europe?

We do not send your data to the companies that created the models. The infrastructure or inference provider receives the request to execute it and strictly follows a zero data retention (ZDR) policy, without storing inference data. At QDivZero, we apply zero logs by default. With Paranoia Mode enabled, we mask detected sensitive data before sending the request to reduce exposure during processing too. HTTPS/TLS protects communications. All three measures work together and also apply when you deploy outside Europe.

How do European ownership and data residency differ?

QDivZero is developed and operated by Valendra Tech SL, a Spanish company, under the European data protection framework. Your selected region determines where each deployment runs. You can choose regions outside Europe: our internal policy of zero logs and no data sent to model providers remains in place. Applicable processing requirements also depend on the region and your application.

Does QDivZero store my prompts, files, or responses?

We apply zero logs to inference content. The only exception is the Responses API, which retains context to continue conversations using previous_response_id. Retention is enabled by default with store: true. Under its privacy policy, QDivZero does not access this content or use it for other purposes, including model training.

Can I disable or delete logs?

Yes. You can delete logs at any time and disable them globally from the platform. Disabling logs globally makes the Responses API unavailable and requests to it fail because it requires retained context. Use Chat Completions instead and send the relevant history from your application.

Does QDivZero use my data to train AI models?

No. Your prompts, responses, and inference data are not used to train, fine-tune, or improve models. The data processed by your application remains yours.

Can I run an LLM privately?

Yes. QDivZero lets you deploy and run LLMs privately on infrastructure prepared for your workload, without configuring and maintaining servers, GPUs, or inference engines yourself.

Can I run Hugging Face models privately?

Yes. You can deploy compatible models from the Hugging Face catalog with QDivZero on private infrastructure. The platform finds compatible capacity, prepares the environment, and makes the model available to your application.

Can I choose where my models run and my data is processed?

Yes. You can select from available regions according to your application's needs, your users, or your organization's data-residency requirements. QDivZero provides infrastructure in Europe, North America, Asia, Oceania, and Africa.

Can I run AI models and LLMs in Europe?

Yes. QDivZero lets you deploy AI models and LLMs on infrastructure available in Europe when you need execution and workload-related processing to remain within the region.

How does QDivZero protect my data during inference?

We encrypt communications over HTTPS/TLS both between your application and QDivZero and between QDivZero and the GPU or public inference provider running the model. This protects your prompts, files and responses in transit, in both directions.

What does zero logs mean in QDivZero?

Inference content is not logged across QDivZero services. The only exception is Responses, which retains context associated with a response id to continue conversations. This information is retained solely for that feature. QDivZero does not access it or use it for other purposes, in accordance with its privacy policy.

Can I use QDivZero to run AI with sensitive data?

QDivZero is designed for workloads that require greater control over their data, with private deployments, region selection, encryption in transit, and inference without content logs. Specific requirements depend on the data type and applicable regulations.

Can I use QDivZero for applications subject to GDPR?

Yes. QDivZero can be used for applications subject to GDPR. You can run models in European regions, control where data is processed, and manage your logs and conversation context. Communications are protected with encryption in transit, and your data is not used to train models. Final GDPR compliance also depends on how your application collects, uses, and manages personal data.