Public education.
Clear, accurate, accessible explanations of how AI works and where it is heading, made for general Hispanic audiences rather than specialists. We meet people on the platforms they already use, in the language they already think in.
Hispanic AI Safety Institute
AI is being built faster than the world can understand it. For 68 million people in the United States, that understanding is not arriving in their language.
HASI is a nonprofit working to make advanced AI safer and more accessible for Hispanic communities, through education, public communication, policy, and research, in Spanish and English.
The Gap
The fastest-moving technology in a generation is least understood where it is needed most.
Hispanic Americans are now the second-largest population group in the country and one of the youngest and fastest-growing. Nearly one in five people in the United States is Latino. Yet the public conversation about artificial intelligence, what it can do, where it is heading, and how it should be governed, is happening almost entirely in English.
The information that does reach Spanish-speaking communities is worse. Spanish-language misinformation is moderated less, and more slowly, than its English equivalent. The result is a measurable knowledge gap that maps directly onto language.
The risk is not only in the information environment; it is in the systems themselves. Frontier AI models are demonstrably less safe outside English. Safety guardrails that hold in English are weaker, or absent, in other languages. As these systems are deployed into hiring, healthcare, education, finance, and immigration, the people least protected by their safety mechanisms are often the people with the least margin for error.
0M
U.S. Latinos in 2024, roughly 1 in 5 Americans and the second-largest population group.
Pew Research Center / U.S. Census, 2024
0%
Share of all U.S. population growth from 2000 to 2024 that came from Latinos.
Pew Research Center, 2025
~30%
Spanish-language misinformation on Facebook flagged with warnings, versus about 70% of comparable English content.
GWU Cisneros Institute, summarizing platform research
11 to 20 pts
How much more likely Latinos who rely on Spanish-language social media are to believe false political narratives.
NYU CSMaP & UC San Diego, PNAS Nexus, 2024
0x
How much more likely lower-resource, non-English prompts are to elicit unsafe AI outputs than high-resource languages in one large study.
Multilingual jailbreak study, arXiv 2310.06474
<5%
Share of one major platform's workforce that was Latino at the time its Spanish moderation gap was measured.
Reported in GWU Cisneros Institute
Our Work
Four lines of work, one mission.
Media and communication.
A standing source of high-quality, factual Spanish-language content on AI risk and governance, built to compete for attention and to push back against the misinformation that currently fills the gap.
Policy and advocacy.
Bringing Hispanic and Latin American perspectives into AI governance, in the United States and across the region, so that the rules being written reflect the communities they will affect.
Research and monitoring.
Tracking how frontier systems behave in Spanish, documenting safety gaps, and surfacing Spanish-language AI risks that English-first research tends to miss.
Theory of Change
Why this work, and why now.
Spanish is one of the most spoken languages on earth and the heritage language of tens of millions of Americans, yet it remains a low-priority language for the people building and governing AI. That mismatch is the opportunity.
Accessible Spanish-language AI safety work is unusually neglected: well-resourced organizations are scarce, and the few credible voices are easily drowned out. It is also unusually tractable: the audiences exist, the platforms exist, and the demand for trustworthy information is already there. And it is unusually leveraged, because the same communities sit at the intersection of U.S. AI policy and Latin American AI governance.
HASI's bet is that a focused institution, building literacy from the ground up while shaping policy from the top down, can move both at once.
Structure
HASI is being established as a dual-entity nonprofit: a 501(c)(3) for education and research, and a 501(c)(4) for policy and advocacy. We are launching through fiscal sponsorship so the work can begin now, while the permanent entities are formed.
Help us close the gap.
We are looking for funders, partners, advisors, and collaborators who believe this work should not wait.
Get in touchhasi-ai.org