13 DIALOGUES · 52 COUNTRIES · SEPTEMBER 2026
Across a year of public conversations, people revealed how AI is becoming a companion, adviser, confidant and a new source of authority. We asked what this changing relationship means for governance — and for being human.
66 PAGES · PDFWHAT WE DISCOVERED
Data
→Algorithm
→Output
Across neuroscience, law, psychology, theology, education, philosophy and product design, different voices repeatedly returned to the same invisible objects: relationships, inferences, dependencies, continuity and the architecture of interaction itself. We call the resulting framework Relational AI Governance.
THE EVIDENCE
Patterns identified independently across disciplines — not derived from theory. Select a finding to explore its core insight and governance implication.
The law protects categories of data. The deeper risk is what systems conclude about a person’s emotions, vulnerabilities, intentions and future behaviour — even from ordinary conversation alone.
Protect mental and neural data according to what it reveals, not where it was collected. Governance must reach the models built about a person and the actions taken from them.
The next frontier of AI governance is no longer only deciding what machines are permitted to do. It is deciding what kinds of relationships we are willing to engineer between humans and machines, and on whose terms.
— REPORT CONCLUSION
FROM EVIDENCE TO ACTION
Protect mind data, including inferences wherever generated
Define what manipulative AI design looks like
Establish duties of relational care across the AI lifecycle
Require the right human institution to remain involved
Measure user outcomes, not only engagement
Design relationships that users can leave
Make AI value formation pluralistic, transparent, reviewable and contestable
Build the evidence that future regulation will depend on
Make regulatory sandboxes genuinely multi-stakeholder
Build existential and spiritual literacy for the AI age
METHOD
The six findings are the author’s qualitative synthesis and should not be read as positions endorsed by every participant.
Thirteen recorded public events across three tracks.
Central claims, audience concerns, gaps and disagreements.
Patterns compared across disciplines and contexts.
Six findings translated into ten recommendations.
READ THE FULL REPORT
Explore the complete evidence, speaker contributions, regulatory gaps, draft principles, recommendations and letters to future generations.