#77: Impact Signals #77 — AI Can Predict the Disaster, Not Fund the Response — Venezuela's Diaspora AI Tools, Mozambique's Drone Pilots, CHAI's Public Health Push
AI for Impact Daily Briefing, July 18, 2026
Top Stories
AI Can Predict the Disaster. It Can't Fund the Response.
Writing in The Energy Mix on July 15, Nakalembe points to the UNFCCC's Fund for Responding to Loss and Damage, the primary international mechanism meant to fund exactly this kind of prearranged disaster response, which has received less than 0.1% of actual need. She cites the return-on-investment case practitioners can use directly: every $1 spent on US disaster mitigation saves $6 in recovery costs, European coastal flood adaptation returns 6-to-1, and global climate-adaptation studies put the 10-year return above $10.50 per dollar invested, roughly 27% average annual return per project. She proposes a rough allocation discipline of $10 spent on local infrastructure and capacity for every $0.01 spent on the AI/digital layer, and points to anticipatory action and planned relocation as proof that institutions can act on a warning instead of just generating one. Why it matters: For NGOs and donors, this is a direct argument for where the next AI-disaster-resilience dollar should go: not into more prediction models, but into the local capacity, staff, legal authority, seawalls, grain reserves, that turns a forecast into a saved life. Cite the loss-and-damage funding gap and the 6:1-to-10.5:1 ROI figures directly in grant applications that pair an AI early-warning proposal with a hard infrastructure budget line.
Venezuela's Earthquake Response Runs on Four Diaspora-Built AI Tools
Following the twin earthquakes that struck northern Venezuela on June 24, at least four separate tools built by diaspora technologists are still running the country's crisis response, Silicon Canals reported July 16. Desaparecidos Terremoto Venezuela, a missing-persons and facial-recognition platform built in three hours with Claude Opus, logged 30,000-plus reports in 48 hours. Samuel Mariña, a Venezuelan developer in California, built Ayuda en Camino, a resource-matching tool with a WhatsApp fallback, in four hours on Replit. Marianne Diaz Hernández, a digital-security expert in Santiago, built the damage-mapping tool terremotove.com in about an hour using Claude Code and Kobo Toolbox. A fourth platform, Somos Acompañamiento, combines missing-persons search with hospital-survivor tracking and has logged more than 84,000 registrations. Three weeks after the earthquakes, these volunteer-built tools remain the primary source of crisis information for affected families, with the Venezuelan state's own response still minimal. Why it matters: This is a replicable pattern, not a one-off: any diaspora community or NGO facing a slow state response now has a demonstrated build time, one to four hours, using Claude, Replit, and Kobo Toolbox, for missing-persons matching, resource-matching, and damage-mapping tools. Note the WhatsApp-fallback design in Ayuda en Camino specifically: it's built for the connectivity conditions of an actual disaster zone, not a demo.
Mozambique Graduates Its First AI-Assisted Drone Pilots
The African Development Bank announced July 17 the completion of a 15-month, $967,000 program, funded through its Korea-Africa Economic Cooperation (KOAFEC) Trust Fund, that trained and certified 30 Mozambican drone pilots and equipped national disaster authorities with nine drones, geospatial mapping tools, and a digital disaster-management system. South Korea's Busan Technopark, Hojung Solution, and Pusan National University led the technical implementation; AI tools sift the drone-collected imagery to measure building damage and search for flood victims from aerial footage. The program wasn't just a pilot exercise: when severe floods hit Mozambique in January and February 2026, the newly trained drone teams ran real aerial damage assessments and supported search-and-rescue, months ahead of this month's formal graduation ceremony in Maputo. Why it matters: This is a template other AfDB member states or KOAFEC-adjacent donors can copy directly: a fixed grant size ($967,000), a defined 15-month training timeline, and a named technical partner bench (Busan Technopark, Hojung Solution, Pusan National University) for any country wanting AI-assisted drone damage-assessment capacity ahead of its next flood season.
CHAI Opens a 2,000-Seat Program to Get Local Health Agencies Using AI Safely
The Coalition for Health AI (CHAI) launched PULSE (Public health Use case and Learning Scaling Engine) on July 16, a national initiative for state, tribal, local, and territorial public health agencies, most of which have lagged well behind federal agencies on generative AI adoption. PULSE will bring roughly 2,000 public health practitioners through five defined use-case tracks; OpenAI and Anthropic have each donated ten enterprise licenses, with up to 2,000 seats available in total. CHAI's 17-member leadership council selects participants and routes them to their preferred use-case committee, where they work with Accenture's AI and public-health specialists to run pilots and turn what they learn into public adoption playbooks. Why it matters: For public-health and social-impact orgs without in-house AI expertise, PULSE is a free, structured on-ramp: enterprise-tier model access from OpenAI and Anthropic, a peer cohort of 2,000 practitioners, and a named technical partner running the pilots, rather than building an AI adoption process from scratch.
Upcoming Events & Opportunities
Kluz Prize for PeaceTech 2026 (Funding)
- Amount: $20,000
- Deadline: July 31, 2026, 11:59 PM EDT
- Eligibility: Tech companies and peace/humanitarian organizations and initiatives using emerging tech to save lives, reduce conflict, or safeguard human rights; priority to young innovators and tech-community members
- Apply: kluzprize.org
Japan Climate Fund — Solar Sharing (Agrivoltaics) Grant Program (Funding)
- Amount: Up to ¥10 million per project (¥20 million total pool this fiscal year)
- Deadline: August 19, 2026
- Eligibility: Agrivoltaics and climate-resilience projects for farmers and local communities, in partnership with local governments or private companies
- Apply: fundsforngos.org
Active Disaster Monitoring (GDACS/OCHA)
- Mexico:** M7.3 earthquake 58 km WSW of Puerto Madero plus a M6.0 aftershock 90 km SW (July 17); USGS significance scores of 973 and 555, yellow and green alert levels; no confirmed casualties as monitoring continues. Source: USGS.
- Philippines:** Mayon Volcano erupting in Albay province, ongoing since at least July 16; GDACS Orange alert, PHIVOLCS Alert Level 3 with 327 rockfalls and a 6 km evacuation radius; thousands evacuated. Source: GDACS.
- Horn of Africa (Ethiopia, Kenya, Somalia):** Regional drought, GDACS Orange alert active since July 3; regional response ongoing. Source: GDACS.
Sources: See individual stories above for full attribution.