Your Data. Your Rules.

At ATLAS, we believe simple: your professional intelligence is your competitive advantage. The relationships you've built, the decisions you've made, the patterns you follow—these belong to you, not us.

When you build your personal world model on ATLAS, the AI learns from your data. But that model is yours. You can export it, delete it, take it somewhere else. We have zero rights to it.

That's not just a policy promise. It's engineered into our architecture.

We've designed ATLAS so it's mathematically impossible for us, our staff, or anyone else to read your data. Here's how:

Your encryption key never leaves your device. When you want us to process something, that key gets encrypted and sent to a hardware-isolated vault. The key is destroyed the moment processing finishes.

Even we can't see what we're processing. Your data is decrypted only inside a secure enclave—a isolated computer chip that has no doors, no back access, and no persistent storage. After processing, everything is wiped.

You can verify this yourself. Our security code is open source. You can read it, audit it, and verify it matches what's running on our servers.

The result: your data is cryptographically inaccessible to ATLAS, our cloud provider, and anyone trying to hack us.

Pro Tier: Your data is encrypted and stored safely with us. We handle backups, sync across devices, and keep everything secure. Perfect for teams.

Ultra Tier: Your data never touches our servers. You keep it on your own infrastructure. We only access it when you ask, then we forget it. This is for executives with the strictest confidentiality requirements.

Choose what fits your threat model.

We operate under strict principles about what data we will not collect, process, or use.

We do not sell your data to advertisers, brokers, competitors, or any third party. This is not negotiable.

We do not use your data to train other users' models or shared foundation models. Your data belongs to you alone. It does not improve ATLAS for other users or train any system beyond your personal model.

We do not perform behavioral tracking across the internet. We do not collect clickstream data, geolocation, or cross-site tracking data.

We do not build marketing profiles from your data. We do not analyze your relationships or decisions to sell access to advertisers or data brokers.

We do not use your data for purposes other than providing ATLAS to you. Every use of your data is constrained to the specific purpose you authorized.

These are not marketing claims—they are enforced through architecture (encryption prevents us from accessing it), contracts (our LLM provider agreement forbids data retention), and operational procedures.

For Pro users, we sometimes ask Claude (made by Anthropic) to help process your data—generating summaries, finding patterns, that kind of thing.

But we have a legally binding contract with Anthropic: your data is never stored, never used to train their models, never used for anything else. It's deleted immediately after the response.

Ultra users run local AI models on your own infrastructure, so Anthropic never sees anything.

You have full control over your data at all times:

- Export: Download all your data in standard formats, whenever you need it.
- Delete: Remove your account, your data, and your trained model permanently.
- Revoke: Immediately stop ATLAS from accessing your data sources.
- Access: Request a complete report of all data we hold about you.
- Visibility: Understand exactly how your information is being processed.

To exercise any of these rights, contact contact@thinkv1.com.

Our complete privacy policy covers the full technical architecture—cryptographic key management, TEE isolation, blind relays, forward secrecy, threat models, and the limitations we are transparent about.

It also includes regulatory compliance details for GDPR, CCPA, HIPAA, and SOC 2.