Skip to main content
Cuadra AI

AI Transparency

We believe you deserve to know exactly how your data is processed and how our AI models work. No legal jargon—just straight answers.

EU-based · GDPR Compliant · No Hidden Surprises

Our Core Promise: Your Data Stays Yours

What makes Cuadra different:

  • Your training data is isolated from all other customers—always
  • We never use your data to train our base models or anyone else's models
  • Third-party AI providers (OpenAI, Anthropic, Google) are contractually prohibited from using your data for training
  • You own everything: your data, your outputs, your trained models

Most AI platforms bury these details in 50-page legal documents. We put them front and center because transparency builds trust.

How Your Data Flows Through Our Platform

📤

1. You Upload Data

Your documents, text, or files are encrypted in transit (TLS 1.3) and stored in isolated databases (AES-256).

🔬

2. We Train Your Model

Our platform processes your data to create custom model parameters—stored separately and linked only to your account.

🤖

3. You Deploy & Use

Your model generates outputs based on your prompts. Only you (and your team) can access it.

🗑️

4. You Control Deletion

Delete your data anytime. Personal accounts: 30-day retention. Business: 90 days for migration.

Who We Work With (And Why You Can Trust Them)

To deliver world-class AI, we partner with leading providers. Here's exactly who sees your data and what they do with it:

AI Model Providers

ProviderWhat They DoData Usage Policy
OpenAI (GPT)Natural language processing & generation✓ Does NOT use API data for training
Anthropic (Claude)Advanced reasoning & text analysis✓ Does NOT use API data for training
Google (Gemini)Multimodal AI capabilities✓ Does NOT use API data for training

Important: When you use our platform, your prompts may be temporarily sent to these providers to generate responses. However, all providers we use have strict "no-training" policies for API data, backed by Data Processing Agreements.

Infrastructure Providers

  • Google Cloud Platform: Hosting, storage, and compute (EU data residency where possible)
  • Pinecone: Vector database for embeddings and similarity search

Business customers: We'll notify you 30 days before adding new providers. You can object or opt-out.

Who Owns What

✅ You Own

  • ✓ Your training data
  • ✓ AI outputs
  • ✓ Trained model parameters
  • ✓ Commercial rights to use outputs

🔧 We Own

  • • The Cuadra AI platform
  • • Base AI models
  • • Training infrastructure
  • • Platform features

Can You Export Your Models?

  • Personal Accounts: You can export training data and outputs, but not model weights (they stay on our infrastructure)
  • Business Accounts: Full model export in ONNX or PyTorch format—take your models anywhere

No attribution required: Use AI outputs commercially without mentioning Cuadra (though we'd appreciate it!).

Being Honest About AI Limitations

⚠️ AI is powerful, but not perfect

Unlike traditional software that follows exact rules, AI models are probabilistic. This means they can:

  • Produce inaccurate outputs ("hallucinations")
  • Show biases from training data
  • Give inconsistent answers to similar questions
  • Fail unexpectedly on edge cases

What This Means for You

Always validate AI outputs before using them in production, especially for:

  • Medical, legal, or financial decisions
  • Safety-critical systems
  • High-stakes business decisions
  • Public-facing content

We're not trying to scare you—we're being realistic. AI is incredibly useful when used responsibly, but it's not a magic solution that works perfectly 100% of the time.

How We Protect Your Data

🔐 Encryption

TLS 1.3 in transit, AES-256 at rest. Keys rotated regularly.

👥 Access Control

Multi-factor authentication, role-based access, audit logs.

🚨 Breach Response

We'll notify you within 72 hours if a breach affects your data.

🇪🇺 GDPR Rights

Access, rectify, delete, or export your data anytime.

What You Can't Do on Cuadra

To protect our community and maintain platform integrity, we prohibit:

  • Illegal content: CSAM, violence, terrorism, hate speech
  • Deceptive deepfakes: Synthetic media without disclosure
  • Unguarded high-risk AI: Medical/legal/financial apps without proper safeguards
  • Model theft: Reverse-engineering our models or other users' data
  • Adversarial attacks: Attempting to break our safety filters

Violations may result in warnings, suspension, or account termination.

Questions About Our AI Practices?

We're here to help. Contact our team for detailed answers about data processing, security, or compliance.

Contact Us