Every time you use a closed-source AI service, you're feeding your data into a system you cannot inspect, cannot audit, and cannot control. You trust the provider's privacy policy—a document that can change without notice and that you almost certainly haven't read.

Open-source AI is not just a technical preference. It's a structural safeguard against the concentration of surveillance power.

The Surveillance Problem

Closed-source AI services collect, store, and potentially analyse every interaction you have with them. This creates a detailed profile of your thinking, your business operations, your creative process, and your decision-making patterns. This data has value—to advertisers, to governments, and to the companies themselves for model training.

The fundamental asymmetry: they know everything about you. You know nothing about their systems.

What Open Source Changes

When AI models are open-source:

  • The code is inspectable: Anyone can verify what the model does and doesn't do with your data
  • The model runs locally: Your data never leaves your machine. No server logs, no training data collection, no third-party access
  • The community audits: Thousands of researchers and developers examine open-source models for hidden behaviours, biases, and vulnerabilities
  • The power is distributed: No single company controls access to the technology

The Current Open-Source Landscape

Open-source AI has reached remarkable capability levels:

  • Meta's Llama models are available in various sizes for local deployment
  • Mistral provides high-performance open models with permissive licensing
  • Google's Gemma offers smaller, efficient models designed for local use
  • Ollama makes running these models locally as simple as a single command

The gap between open-source and proprietary models is narrowing. For many professional tasks—drafting, analysis, brainstorming, code generation—open-source models running locally are more than adequate.

The Honest Trade-Off

Open-source AI is not universally superior:

  • Frontier capabilities (the very latest, most capable models) are typically available as closed-source services first
  • Convenience is lower: local setup requires more technical knowledge than signing into a web interface
  • Compute requirements for the largest open models exceed what consumer hardware can handle
  • Support is community-driven rather than commercially backed

The pragmatic approach: use open-source/local AI for sensitive operations and closed-source services (with paid tiers and privacy guarantees) for non-sensitive tasks that benefit from frontier capabilities.

The Broader Stakes

This isn't just about individual privacy. It's about the architecture of power in an AI-mediated world:

  • If AI is controlled by a handful of corporations, those corporations become gatekeepers of intelligence itself
  • If AI is open-source and distributed, intelligence becomes a public utility that anyone can access and deploy
  • If governments can compel closed-source AI providers to share user data, every AI interaction becomes a surveillance vector
  • If AI runs locally on your hardware, government access requires physical seizure of your device—a much higher legal bar

Supporting open-source AI is an investment in a future where intelligence tools are distributed rather than centralised. That distribution is the structural safeguard against surveillance overreach.

What You Can Do

  1. Run local models for your sensitive work (Ollama + any open model)
  2. Support open-source AI projects (use them, contribute feedback, fund development)
  3. Advocate for open-source in your professional network and content
  4. Stay informed about open-source developments and contribute to the community conversation