Recommendations of the Global AI Center, in the spirit of the Artificial Intelligence Constitution

At the heart of the AI Constitution lies the conviction that no one voice, human or artificial, may speak alone for what humanity holds in common. Applied to information integrity, this means that truth in the digital space cannot be monopolised by a single state, platform, or algorithm, and that artificial intelligence must never become an unauthorised spokesperson for civilisation. From this principle, the Centre offers the following recommendations:

 

📌AI systems should acknowledge the limits of their knowledge. Generative AI should be designed to cite verifiable sources and to state openly when no reliable information exists, rather than produce a confident but unfounded answer. Honesty about uncertainty should be recognised as a core safety standard, not an optional feature.

 

📌Training data should be protected against deliberate poisoning. Developers should apply source-provenance safeguards and independent audits so that coordinated campaigns of fabricated content cannot distort what AI systems present as fact, particularly in smaller linguistic and cultural contexts that are easier to target.

 

📌Decisions affecting freedom of expression must keep a human in the loop. Automated moderation should never be the final word on removing lawful speech. Every such decision should be explainable, subject to independent appeal, and accompanied by clear notice to the user.

 

📌Every language deserves equal protection. Safeguards, terms of service and reporting mechanisms should be available in local languages, so that communities speaking less widely used languages are not left more exposed to manipulation than others.

 

📌Measures against disinformation must remain legal, necessary and proportionate. The fight against disinformation must never become a pretext for silencing journalists, human rights defenders, or dissenting voices. Integrity and freedom are partners, not rivals.

 

📌Information integrity is part of a healthy digital ecosystem. In line with the Centre's Eco-Cyber Security doctrine, the information space should be treated as a shared environment requiring stewardship, resilience, and long-term care, in the same way we care for the natural world.

 

📌Media and AI literacy should begin early. Education systems should equip young people not only to recognise false information, but also to understand how AI systems generate content and why they can be wrong.

 

The Global AI Center POLLYPRIANY expresses its sincere gratitude to the United Nations Office for Digital and Emerging Technologies and all partner organisations for this open and inclusive dialogue, and reaffirms its commitment to contributing to the implementation of the Global Digital Compact in a spirit of multilateral cooperation and human dignity.