Where Your Voice Goes: What is Processed on the Device and What in the Cloud

When choosing an AI assistant for negotiations, the main question isn't "how smart is it?" but "where does my voice go?" For a doctor, lawyer, financier, or negotiator, this isn't about convenience—it's about law and reputation: medical and attorney-client privilege, corporate compliance, and sensitive deals.

That's why we want to explain honestly how data processing works in SpotMax—without any marketing spin.

What runs on your device

Speech recognition and diarization (speaker separation) run locally on your computer. You can keep them entirely on your device, meaning the audio never leaves your machine. This isn't just a checkbox in the settings; it's how the product is built by default.

What can go to the cloud — and only by your choice

Certain features can run in the cloud: translation of remarks and, optionally, cloud processing mode—if your hardware is too weak to run local models at high quality. This is the only data that can ever leave your device, and you have to enable it yourself. If you choose not to, the entire pipeline remains strictly on your machine.

We are upfront about this because the difference between "fully local" and "partially cloud-based by choice" is fundamental, and blurring the lines would be dishonest.

No central database, no backdoor

Here is the key difference. Cloud-based meeting assistants funnel the conversations of all users into a single data center. While convenient for the provider, this is also the ultimate vulnerability: a centralized database of conversations is both a single point of failure and an inherent backdoor.

A single leak, one regulatory subpoena, or one compromised employee, and it's not just your call that gets exposed—it's everyone's. The biggest backdoor in any system isn't a clever piece of malicious code; it's the very existence of a place where everyone's data is stored together.

SpotMax has no such place. We do not collect your conversations into a shared database: there is nothing to seize with a single warrant, nothing to leak en masse, and no master key to hand over. Here, privacy is a feature of the architecture, not just a promise in a privacy policy.

What we don't do

We do not track or profile you, we do not sell your data, and we do not train models on your conversations. There is no background analytics "to improve the product" that actually just serves as a cover for data harvesting.

Where we are headed

A fully local pipeline—speech recognition, translation, and real-time prompts running on your device without any server involvement—is our ultimate goal and development roadmap, rather than our current state. This is exactly why we invest in local models and fund bounties for their improvement: the better these models run on consumer-grade hardware, the less reason there is to ever touch the cloud.

Why this matters now

Virtually all popular AI meeting assistants upload your recordings to their cloud and store them there. This represents a fundamentally different philosophy: your negotiations become part of someone else's infrastructure, forcing you to trust them with your security. SpotMax is built on the opposite principle: sensitive conversations should stay exactly where they took place—with you.