Back to Home

About ThisChatNeverExisted

A privacy-first AI chat application where your conversations truly remain private - not because we promise, but because it's architecturally impossible for them to leave your device.

What is ThisChatNeverExisted?

ThisChatNeverExisted is a chat application that processes AI conversations entirely on your device using Chrome's built-in Prompt API with Gemini Nano. Unlike traditional AI chat services that send your messages to remote servers, every word you type and every response you receive stays completely local.

Your conversations exist only in your browser's RAM - no databases, no cloud storage, no backend servers. When you close the tab or trigger the burn ritual, they're gone forever.

Key Differentiator: Privacy Through Impossibility

Most chat applications ask you to trust their privacy policy. We make privacy violations technically impossible:

  • No backend to compromise - Static HTML/CSS/JS hosted on a CDN, zero server-side processing
  • No database to breach - Conversations live in React state (RAM only), cleared on page close
  • No network calls during chat - All AI processing happens locally using Chrome's on-device model
  • Open source verification - Every line of code is publicly auditable

This isn't a privacy promise - it's an architectural guarantee. See our Privacy Architecture for technical details.

Signature Feature: The Burn Ritual

Privacy in most applications is abstract and invisible. We make it visceral. The burn ritual transforms data destruction into a tangible experience - watch your conversation turn to ash and fade away, making the ephemerality of your data emotionally real.

It's not just about deleting data (which happens automatically anyway when you close the tab). It's about feeling the privacy you have, making the invisible visible.

Who is this for?

ThisChatNeverExisted is designed for people who need genuinely private AI conversations:

  • Privacy-conscious individuals - People who understand that “encrypted in transit” still means your data reaches someone else's servers
  • Sensitive questions - Medical concerns, legal questions, personal dilemmas, financial planning - topics you wouldn't want in a company's training dataset
  • Creative professionals - Writers, designers, strategists protecting intellectual property and early-stage ideas
  • Technical skeptics - Developers and security professionals who want to verify privacy claims themselves

Frequently Asked Questions

Why Chrome Desktop only?

ThisChatNeverExisted uses Chrome's Prompt API, an experimental feature (being standardized) that provides access to Gemini Nano, a compact AI model that runs entirely on your device. As of Chrome 138+, this API is Chrome-exclusive and only available on desktop platforms (Windows 10/11, macOS 13+, Linux).

We chose this approach specifically for privacy - no third-party API keys, no remote model calls, no data transmission. The trade-off is platform limitations, but the privacy guarantee is worth it.

Supported: Chrome Desktop 138+ (Windows 10/11, macOS 13+, Linux)
Not supported: Mobile browsers, Safari, Firefox, Edge (for now)

Is the 22GB download required?

Yes, but it's a one-time download that stays on your device. The Gemini Nano model is approximately 22GB and must be downloaded to enable local AI processing. This happens through Chrome's built-in model management system.

Think of it like installing a high-quality offline application - the upfront cost is significant, but you never need to download it again. The model persists across browser sessions and is shared by all sites using the Prompt API.

Why so large? Language models require billions of parameters to understand and generate human-like text. 22GB is actually compact for an AI model - most cloud-based models are 100GB+.

Can I export conversations?

No, by design. Adding export functionality would contradict our core privacy promise. Here's why:

  • Export requires saving to disk (violating RAM-only architecture)
  • Exported files can be accidentally synced to cloud storage (Dropbox, iCloud, etc.)
  • Exported files create a permanent record we promised wouldn't exist

If you need to save information from a conversation, we recommend manually copying specific insights you want to keep. This forces conscious decisions about what's worth preserving, rather than saving everything by default.

Is this really free? Forever?

Yes, completely free. No subscriptions, no API costs, no hidden fees, no “freemium” upsells.

Here's the business model: there isn't one. This project exists because:

  • No backend = no server costs
  • Static site hosting is essentially free (Cloudflare Pages CDN)
  • AI runs on your device = no API costs
  • Open source = community-driven development

The only ongoing cost is the domain name (~$10/year). This isn't a startup looking for an exit - it's a privacy tool that will work exactly the same way in 5 years.

Can you add [feature X]?

Open source means you can! ThisChatNeverExisted is licensed under MIT, which means you can:

  • Fork the repository and add features yourself
  • Submit pull requests with improvements
  • Create your own version with different features
  • Check the GitHub repository for roadmap and contribution guidelines

That said, we're selective about accepting features that might compromise the privacy-first architecture. No feature is worth sacrificing the core guarantee.

What if Chrome deprecates the Prompt API?

We'll pivot to another local LLM solution. The Chrome Prompt API is currently the most accessible way to run AI locally in a browser, but it's not the only way.

Backup options include:

  • Ollama integration - Run local models through a desktop application
  • WebLLM - Browser-based model loading using WebGPU
  • LM Studio - Local model hosting with browser bridge
  • Browser AI API - Emerging standard for on-device AI (when ready)

The core architectural principle (no backend, no data transmission) will remain unchanged. Only the technical implementation of local AI would change.

Roadmap

We're MVP-first. The current version focuses on proving the core concept: private, local AI chat that people actually want to use. Future enhancements depend on feedback and community interest.

Planned Features (Post-MVP)

These are ideas, not promises. We'll prioritize based on what users actually need:

  • Multiple burn animations - Different visual styles for the burn ritual (embers, dissolve, shred)
  • Role-play modes - Pre-configured conversation styles (coding assistant, brainstorming partner, etc.)
  • Model options - Support for different local AI models as they become available
  • Accessibility improvements - Enhanced keyboard navigation, screen reader support, high-contrast themes
  • Browser extension - Bring private AI assistance to any web page

What we'll never add:

  • Cloud sync or conversation backup (violates privacy architecture)
  • User accounts or authentication (creates tracking surface)
  • Analytics on conversation content (architecturally impossible anyway)
  • Third-party integrations that require data transmission

Get Started

Ready to try truly private AI chat? Here's what you need:

  1. Chrome Desktop 138+ (Windows 10/11, macOS 13+, or Linux)
  2. 22GB free disk space (one-time model download)
  3. 5 minutes for model download (varies by connection speed)

Start chatting or learn how it works.

Want to verify our privacy claims? Follow our step-by-step Verification Guide.

Questions or feedback? Visit our GitHub Issues or check the repository README.