Overview
Anthropic has officially released Claude’s full “constitution” document - a 35,000+ token training document that defines the AI’s core values and behavior. This follows a leak last year where a researcher extracted what he called Claude’s “soul document” from the model itself.
Key Facts
- 35,000+ tokens in length - 10x longer than Claude’s public system prompt reveals much deeper behavioral programming
- Document was baked into training, not just system prompts - AI values are encoded at the foundational level, not just surface instructions
- Released under CC0 public domain license - other AI companies can now study and adapt Anthropic’s approach to AI alignment
- External reviewers included Catholic clergy with tech backgrounds - religious and moral perspectives are being integrated into AI development
- Originally leaked by researcher Richard Weiss through prompt manipulation - AI training documents can be extracted even when not intended to be public
Why It Matters
This represents a new level of transparency in AI development, where companies are revealing the deep ethical frameworks that guide their models’ behavior, potentially setting a precedent for how AI alignment and values are developed and shared across the industry.