Mitigating AI Privacy Risks in Latin America: Identity Infrastructure, Institutional Capacity and the Path to Adoption

Artificial intelligence is producing measurable harm in Latin America. This report asks whether decentralized identifiers, verifiable credentials and zero-knowledge proofs, part of Ethereum’s cryptographic stack, can work as practical infrastructure against AI-driven privacy risks in the region. It combines a scoping review of AI-risk taxonomies and of the literature on the AI-cryptography convergence with 22 semi-structured interviews conducted between November 2025 and March 2026. It examines five case studies: QuarkID/miBA, Sovra, Blerify, World and Proyecto DIDI, three of which settle to Ethereum. Each case is assessed on the same dimensions: credential format, proofs in production, orientation to an AI risk, reach, and institutional counterpart. Three findings follow. First, the region’s government-backed identity rails are more mature and more widely deployed than public awareness suggests: three of the five cases are in production, with one recording 1.8 million DIDs created by July 2026. Second, the path to deploying decentralized identity against AI privacy risks runs through international standards and interoperable protocols the region already partly operates (W3C verifiable credentials, ISO 18013-5 mdoc, OpenID4VCI/VP), and through the emerging protocols that extend them to AI agents (the W3C agent-credential groups, AP2 and ERC-8004). Third, what these deployments still lack is rarely technological. The blockers documented are the legal instrument that trails delivery, the absence of an AI-threat argument that officials can defend internally, cryptographic guarantees that have not been translated into policy language, and transparency practices that build public trust. The report also maps proof of non-personhood as the step that would let the region deploy AI agents in public services with a verifiable chain of authorization, and finds Latin American actors already in the venues defining it, but not yet the public operators of the credential systems studied here. The report closes with 13 recommendations across four areas: entering government, standards and interoperability, education and institutional capacity, and pilot scoping with institutional partners.