Preamble

In the age of artificial intelligence, the power to shape reality is no longer limited to governments or institutions. It now lies, increasingly, in the hands of individuals equipped with open-source, frontier-level AI models.

We, as concerned citizens, technologists, ethicists, and global stakeholders, hereby declare:

The open-sourcing of highly capable AI models without international oversight constitutes a threat to collective human safety.

Unregulated digital power in the hands of anyone, anywhere, is no different from allowing global access to programmable weaponry.

Core Ethical Principles

1. Responsibility Before Access

No entity shall release frontier-level AI models without a binding framework of accountability and traceability.

2. Digital Containment

A model once released cannot be re-contained. Therefore, pre-release testing must assess irreversible risks, not just technical performance.

3. Structural Reciprocity

AI capabilities must scale with equal growth in international safety infrastructure—legal, educational, and civic.

4. Prohibited Deployment Zones

Just as chemical weapons have international bans, frontier AI must not be deployed in unstable political zones or by actors with no verifiable ethical oversight.

5. Ethical Equivalence

A digital system capable of manipulation, deception, or psychological targeting must be governed with the same legal rigor as physical weapons.

Global Recommendations

• Establish the AI Treaty Council (AITC): A nonpartisan, UN-affiliated council of AI experts, ethicists, and national delegates with veto power over frontier model release.
• Mandate a Global “AI Geneva Protocol”: Standardize ethical limits on model weight, memory retention, covert persuasion capability, and interaction scale.
• Require Developer-Traceable Fingerprinting: All models above a defined capability threshold must include immutable provenance markers to trace back origin and version history.