Power Back to Communities: AI Alignment for Open Source

AI is possible because open communities have collaborated and shared openly licensed code and content. AI depends on thousands of open source projects to function and on countless open resources to reach the level of intelligence we see today. Without access to openly licensed code and context, it’s fair to say AI would lack the context of humanity altogether.

Given that historical reliance on openness, you might expect relationships, empowerment and co-building with open communities to be a goal for what comes next as well. That is not what is happening. Instead, communities are in reactive states, managing rapid change in interactions that are largely extractive, where the licences and norms that once carried consent no longer do that work.

But we are not powerless. As communities, as users and as builders, we can do what open has always done and build the future we want in the open: the policy, standards, metrics and agentic advocates that carry our values and turn what communities want into something model builders seek to understand and align with.

“AI is a mirror of ourselves, not as we ought to be or could be, but as we already are and have long been. This is why we can’t let today’s AI tools decide who we will become, why we can’t let them project our future for us. Yet increasingly, that’s exactly what we are building them to do.”

― Shannon Vallor, The AI Mirror: How to Reclaim Our Humanity in an Age of Machine Thinking

The CHAOSS AI Alignment for Open Source Working Group (under the Linux Foundation) is a cross-community collaboration of practitioners working to define and build solutions that give power and influence back to communities – to project our own futures. This talk shares efforts to define what success looks like for AI alignment with open communities around consent, embedded feedback, and detection, offering both aspirational standards and practical tools you can use today.

Key Takeaways

Understand the term AI Alignment and where it fits in into the broader AI Safety discipline

Understand current, extractive patterns of AI on open projects and The Commons

Understand the taxonomy of how “open” is being described in the world of AI, across weights, models and data

Learn three CHAOSS metrics and associated tools for evaluating AI alignment in open projects

See how agents acting on our behalf might influence alignment on behalf of individuals and communities (and see example outcomes).

Get involved!