Privacy & Data Sovereignty

Your Data. Your Machine. Your Privacy.

We believe autonomous AI should empower you without compromising your personal privacy or system security. Read how Yaazh protects your data with local-first engineering.

LAST UPDATED: SEPTEMBER 2026•VERSION 2.5 STABLE

100% On-Device & Zero Telemetry

No remote tracking servers, no usage analytics, and no behavioral surveillance. All process executions and decisions stay local to your computer.

Encrypted API Key Security

Your model credentials (Gemini, Claude, OpenAI) remain securely inside your local .env file. Keys are masked in the UI and stripped from audit logs.

Ephemeral Visual Buffer

Screenshots are captured strictly while a user-initiated task is running, held in temporary memory, and discarded immediately upon completion.

Full Data Sovereignty & Export

All transcripts and action logs are saved locally in your own directory. You can inspect, export, or permanently erase records at any time.

01. ARCHITECTURE

Local Execution & Infrastructure

Yaazh AI is architected from the ground up on a strict Local-First foundation. When you run the Yaazh Desktop Agent, Web Tools, or Chrome Extension, our software interacts directly with your operating system, local browser, and target applications.

We operate no centralized telemetry or logging servers. We never intercept, inspect, or monetize the commands you issue or the files you manipulate.

02. CREDENTIALS

API Keys & Direct Provider Communication

When you configure model providers (such as Google Gemini, Anthropic Claude, OpenAI, or local Ollama instances), your API keys remain stored exclusively on your local file system within your encrypted configuration files.

  • Direct TLS Encryption: Requests travel directly from your computer to the AI provider endpoint via HTTPS/TLS 1.3 without intermediary proxies.
  • Log Scrubbing: Secret keys and sensitive credentials are automatically sanitized from in-app logs and diagnostics.
  • Offline Model Support: You can connect local LLM instances (Ollama / vLLM) for complete air-gapped, zero-cloud execution.
03. SCREEN VISION

Screen Captures & Vision Safety

To click buttons, navigate portals, and fill forms, Yaazh captures visual screen coordinates. We adhere to rigorous visual privacy safeguards:

Zero Cloud Screenshot Archival

Visual frames exist in temporary system RAM only for the duration of the active visual calculation. Frames are instantly discarded once action coordinates are computed and are never saved to external cloud databases.

04. AUDIO

Voice Commands & Audio Privacy

When you use voice mode in Tamil, English, or Tanglish, microphone audio is passed to your configured Speech-to-Text engine (such as local Whisper models).

Audio streams are never continuously monitored or passively recorded in the background. The microphone is active only while you hold the Push-to-Talk button or activate recording (Ctrl + Space).

05. USER RIGHTS

Data Sovereignty, Export & Deletion

You retain 100% ownership and sovereignty over your data:

AUDIT TRANSPARENCY

Every action, keystroke, and step executed by the agent is recorded in local audit trails accessible via Ctrl + H.

INSTANT PURGE

You can delete individual conversations or wipe your entire history at any time with a single click in Settings.

Privacy Inquiries & Legal

Have questions regarding our privacy architecture?

Our security and engineering team is here to provide full transparency on data flow, zero-retention policies, and enterprise on-premises deployments.

Get Started