Brazilian civil society leader says AI governance must recognize community 'trust infrastructures'

A commentary argues that global AI governance must recognise the trust structures already operating inside communities. It starts from Geneva, which hosted the opening session of the UN dialogue on AI governance and where Secretary-General António Guterres warned that computing power, data and talent sit with a handful of companies and countries. The author says the Dialogue reaches organisations able to take part in multilateral processes, but not quilombola communities in Rio de Janeiro or low-income families in Albuquerque, and calls this a structural trust deficit.

Two bodies of research support the case. Discussions with residents of Denver, Atlanta, New Mexico and Appalachia fed the report "The Blueprint for Equitable Digital Participation", whose participants did not reject technology; they preferred to learn about it at libraries and through networks they trusted locally. In Brazil, the Territórios Digitais research, carried out in a quilombo and an Indigenous village, found that trust comes from territorial networks of leaders, educators, health workers and especially women community leaders. The author calls the resulting role sociotechnical mediation. One example is a participant who used generative AI to read a bill threatening his community's rights but checked the result with a trusted partner. Another is a quilombo leader threatened after a false story spread on messaging apps.

The piece also cites an early report from the UN's scientific panel on AI, which says the AI divide is about capacity to influence development, notes that 118 countries are not engaged in major governance discussions, and admits that the path from personal AI use to social outcomes is poorly understood.

The author's three principles: treat territorial trust networks as implementation infrastructure, fund community mediation permanently as a public good, and treat long-term territorial research as primary evidence.