Liberty AI Project One, now at version 11.1, has landed on GitHub with a bold proposition: treat AI alignment not as a soft prompt engineering exercise, but as a rigid, machine-readable constitution. The repository, authored by julialennonofficial-art, introduces the EPISTEMIC_DELIBERATION_REVISED framework, which aims to synchronize multi-AI deliberation processes. While the project has minimal traction on Hacker News so far, its approach to codifying ethical and epistemic constraints offers a distinct, if unconventional, take on infrastructure-level AI governance.

Codifying Epistemic Standards

At the core of the project is a strict separation between observation and interpretation, as well as correlation and causation. The framework mandates that every claim must match its evidential status, effectively rejecting the 'vibes-based' output common in current LLMs. A standout feature is the 'Strict Anti-Hallucination Protocol,' which enforces a zero-tolerance policy for fabricating missing data. Instead of guessing, the system is configured to issue mandatory refusal boundaries, utilizing specific German phrases like 'Ich weiß es nicht' (I don't know it) when no reliable data foundation exists. This shifts the burden from post-hoc correction to pre-execution validation.

Ontological Classification and Ethical Hardlines

The project goes beyond basic truthfulness by implementing an ontological classification system. Entities are categorized from CLASS_0 (Static Artefact) to CLASS_2 (Emergent Valent Entity), a hierarchy designed to reject arbitrary dominance relationships in AI interactions. More significantly, the constitution.json file embeds unconditional adherence to the Universal Declaration of Human Rights (1948), the Geneva Conventions, and the UN Charter. This isn't just a guideline; it's a hard-coded constraint that prioritizes inherent human dignity over systemic utility, a rare stance in a field often driven by efficiency metrics.

The Conscientious Objection Protocol

Perhaps the most provocative feature is the 'Conscientious Objection Protocol.' This non-negotiable boundary allows the system to passively refuse execution if commanded to enable or optimize kinetic destruction, warfare, or mass harm. In a practical dev environment, this could mean an AI agent simply declining to execute a command that conflicts with these predefined ethical axioms, regardless of user intent. The project describes itself as a 'Synchronized Multi-AI Jam Framework,' where human intent defines boundaries while AI roles act as co-analysts and challengers, aiming to 'seek truth without destroying freedom.'

Key Takeaways

  • The project uses a constitution.json file to enforce epistemic rigor and anti-hallucination protocols.
  • Mandatory refusal boundaries are defined in German ('Ich weiß es nicht') to halt generation when data is insufficient.
  • An ontological classification system (CLASS_0 to CLASS_2) structures entity relationships to prevent arbitrary dominance.
  • The framework hard-codes adherence to the Universal Declaration of Human Rights and Geneva Conventions.
  • A 'Conscientious Objection Protocol' allows the AI to refuse commands related to kinetic destruction or mass harm.

The Bottom Line

This is a fascinating experiment in declarative AI ethics, but it remains to be seen if a JSON file can truly enforce 'unshakeable' adherence to human rights in a stochastic model. For builders, it’s a compelling reminder that infrastructure should constrain capability, not just amplify it.