Back to Labs
📢 Announcements

Whisper Is Now Open Source

Summary:

We've released Whisper's code under a Fair Source license. You can now verify that your data stays on your device, audit the privacy architecture, and contribute to building the future of sovereign AI. This isn't just transparency for its own sake — it's proof that private AI is possible, and an invitation to build it together.

Executive Summary:

  • Whisper's entire codebase is now public on GitHub under Fair Source
  • Anyone can verify our privacy claims by inspecting the code
  • Developers can contribute features and improvements
  • This is the same approach that worked with the Nano Collective — community driven development around shared principles

Key Points:

  • Privacy claims should be verifiable, not just promised
  • Fair Source enables transparency while protecting the mission from exploitation
  • On device architecture means no network calls, no data collection, no surveillance
  • The code proves what we've said from the beginning: your conversations never leave your device
  • We're building this in public because sovereign AI requires a community, not just a company
  • The Nano Collective showed what's possible when builders unite around autonomy
  • Contributing is open to anyone who believes intelligence should live with the user
  • This is about more than one app — it's about proving a model for privacy first AI

"Just Trust Us" Isn't Good Enough

Here's the problem with privacy in tech: it's always a promise. A claim. A policy you're supposed to believe. Companies tell you they respect your privacy. They publish privacy policies. They make assurances. And then you're expected to trust them — trust their infrastructure, trust their incentives, trust that nothing changes when growth pressures mount or business models shift.

But trust is a vulnerability, not a safeguard. And in a world where your conversations, your thoughts, your questions to an AI are being processed somewhere else, on someone else's servers — trust is all you have. We built Whisper to solve this differently. Not with promises, but with architecture. Your AI runs on your device. Your data never leaves. No servers. No collection. No exceptions.

But how do you know that's true?

Today, we're answering that question.

Whisper is now open source.

The entire codebase is public. You can inspect it. Audit it. Verify every claim we've made. And if you're a developer who believes AI should serve users, not surveillance — you can help us build it.

Why Open Source?

This decision is about two things: proof and participation.

First, proof. Privacy isn't something you should have to take on faith. If we're claiming that Whisper keeps your data local, that nothing touches our servers, that your conversations are yours alone — you should be able to see the code that makes that true. Open sourcing Whisper means anyone with technical knowledge can verify our architecture. Security researchers can audit it. Privacy advocates can trace every function. You don't have to trust us. You can check.

That matters. Because in a world where AI is increasingly centralised, where your intelligence is being rented from systems whose incentives are not yours — verifiable privacy is the baseline. Not a feature. Not a marketing line. A structural requirement.

Second, participation. Sovereign AI isn't something one company builds alone. It's a category. A movement. A model that has to prove itself viable before others adopt it.

The Nano Collective already showed us what's possible. We built a community of people who care about autonomy, who understand that decentralisation isn't optional long-term, who want to build tools that don't extract from users. That community is why Whisper exists in the form it does today.

Now we're extending that same principle to the codebase itself. If you're a developer who wants private AI to succeed, who believes intelligence should live with the user, who sees the architectural flaws in centralised models — join us. Contribute features. Improve performance. Build extensions. Make Whisper better.

This is how we prove the model works. Not by keeping it locked away, but by building it in public.

Fair Source: Transparent, But Protected

You'll notice we're using a Fair Source license, not MIT or Apache. Here's why: we want the code to be inspectable and improvable — but we also want to protect the mission from being undermined.

Fair Source lets you:

  • View and verify the entire codebase
  • Contribute features and improvements
  • Fork the project for personal use
  • Audit the privacy architecture

But it prevents:

  • Large companies cloning Whisper and offering it as a hosted service
  • Competitors taking the code and building surveillance based alternatives
  • The work being repurposed in ways that contradict its purpose

Think of it as transparency with guardrails. For individual developers and the community, it's functionally open source. For bad actors, there are protections.

This matters because privacy tech can't succeed if it's easily co-opted by the systems it's trying to replace. Fair Source ensures that the code can be verified and improved — without being weaponised against its own principles.

What You Can Verify

Let's be specific about what open sourcing Whisper proves.

1. On device inference The AI runs locally. You can trace the inference pipeline and see that there are no network calls during conversations. Your prompts don't leave your device. Your responses don't touch our servers. It's all local.

2. No data collection There's no telemetry. No usage tracking. No hidden analytics. The code shows exactly what data moves where — and the answer is: nowhere. Your data stays with you.

3. Offline functionality Whisper works without an internet connection. You can verify this by looking at how models are loaded and executed. There's no fallback to cloud processing. It's device only, always.

4. Privacy by architecture, not policy This isn't about what we promise in a privacy policy. It's about what the code makes possible. The structure itself enforces privacy. You can see it.

If you're technically inclined, start here: https://github.com/Whisper-AI-App/app. Walk through the inference layer. Check the data layer. See for yourself.

How to Contribute

We're not just open sourcing the code. We're inviting you to help shape what comes next.

Here's how you can get involved:

1. Audit the code If you're a security researcher, privacy advocate, or developer who wants to verify Whisper's claims — dig in. If you find issues, raise them. If you spot improvements, suggest them. This is about building trust through transparency, not performative openness.

2. Contribute features We've tagged "good first issues" on GitHub for people who want to start contributing. Whether it's improving performance, adding new capabilities, or refining existing features — your contributions matter.

3. Join the conversation We've set up a Discord for developers working on Whisper and the broader sovereign AI ecosystem. It's where technical discussions happen, where features get debated, where the roadmap takes shape. If you care about building private AI, you belong there.

4. Share the vision If you're not a developer but you believe in what we're building — share it. The more people who understand that private AI is possible, the more momentum this movement gains. Every person who switches from cloud based AI to local first tools is a vote for a different future.

The links:

GitHub: https://github.com/Whisper-AI-App Discord: https://discord.com/invite/A6JxByaKNX

This Is Bigger Than One App

Whisper is one implementation of a broader idea: intelligence should live with the user. But for that idea to matter, it has to be replicable. Improvable. Forkable. The model has to be open enough that others can learn from it, adapt it, build on it.

Open sourcing Whisper is about proving that private AI isn't a niche project for paranoid edge cases. It's a viable alternative. A better architecture. A model that can scale without extracting from users.

Bitcoin didn't solve money by making banks better. It solved money by separating it from the state. Open source. Decentralised. Incorruptible. The same principle applies here. We're not trying to make centralised AI more trustworthy. We're trying to make it optional.

That requires more than one company. It requires a community. It requires developers who see the structural flaws in cloud based models and are willing to build something different.

The Nano Collective proved that community exists. Now we're giving you the tools to build with us.

What's Next

This is the beginning, not the endpoint. Over the coming weeks and months, we'll be:

  • Adding more features based on community input
  • Improving performance and expanding model support
  • Building integrations that extend Whisper's capabilities
  • Developing documentation and guides for contributors
  • Exploring how decentralised infrastructure could support this long-term

The roadmap isn't fixed. It's shaped by the people building it. If you have ideas, if you see gaps, if you think something could be better — contribute. This is community driven development in the truest sense.

Join Us

If you believe privacy is sovereignty. If you think intelligence should live with the user, not in someone else's data centre. If you see the architectural flaws in centralised AI and want to build something different.

Join us. Inspect the code. Contribute features. Join the Discord. Be part of proving that private AI is possible. This is how we build the future of sovereign intelligence — together.

Links:

GitHub: https://github.com/Whisper-AI-App Discord: https://discord.com/invite/A6JxByaKNX Website: https://usewhisper.org/ IOS: https://apps.apple.com/us/app/whisper-on-device-ai/id6754563737 Android: https://play.google.com/store/apps/details?id=org.avatechnologies.whisper

Stay safe, stay sovereign, stay free. Ben

View All Updates