Humanitarian AI Today

David Schoeller-Diaz, Impact Engagement Manager formerly with the CyberPeace Institute, joins Humanitarian AI Today producer, Brent Phillips, to discuss the urgent need to conceptualize cybersecurity and artificial intelligence as public goods rather than individual responsibilities. Drawing from twenty years of experience in the humanitarian and peace-building sectors, David highlights how grassroots organizations are currently targeted by sophisticated threats like ransomware and spyware without adequate systemic support. He advocates for a shift toward collective infrastructure, such as regional cybersecurity operations centers and mandatory platform responsibility, to protect the entire social fabric from the cascading impacts of cyberattacks.

The conversation further explores how community engagement principles from David's work with UNICEF, including localization and building on local capacity, can be applied to AI governance. David warns against the dangers of corporate dependency and "parachuting" external experts into aid and development contexts, instead proposing a "community immunity system" powered by federated learning and blockchain to share threat intelligence while maintaining data sovereignty. Both speakers emphasize that authentic AI literacy and genuine partnerships with local technologists are essential to ensuring that emerging technologies support democratic civic spaces rather than reinforcing existing vulnerabilities.

Thomas Byrnes, CEO and Lead Consultant of MarketImpact, speaks with Brent Phillips, Humanitarian AI Today producer, about the growing challenge of "shadow AI" in the humanitarian sector. Byrnes defines shadow AI as the unauthorized use of AI tools like ChatGPT and Gemini by organization staff and field teams. While these tools offer benefits and productivity gains, their unofficial use introduces serious risks, such as the accidental sharing of sensitive beneficiary data, linguistic errors, poor decision making in aid and protection contexts, and "hallucinations" that might lead to incorrect assessments during program design.

To mitigate these risks, Byrnes introduced AidGPT, a training brand that provides humanitarian teams with the skills to use AI tools ethically and safely. A central component of this training is the AI workflow card, an open-source toolkit designed to force users to pause and define the AI's role, set guardrails against misinformation, and verify outputs against approved data checklists. Byrnes emphasizes that while AI tools can provide much-needed efficiency in an underfunded sector, they must be treated like "junior staff members" who require constant supervision and human accountability. Ultimately, he argues for breaking the stigma and "shame" surrounding AI use to facilitate an open, community-wide dialogue about risks, operational best practices and lessons learned.

In Humanitarian AI Today’s first interview recorded in Spanish, David Alejamdro Schoeller-Diaz guest hosts a special interview with Catalina Rebollo, a journalist and regular contributor to Wired magazine in Spanish covering technology and artificial intelligence. Catalina specializes in in-depth reporting on news and complex issues with a particular focus on human rights, inequity and disinformation. She recently wrote on humanitarian technology under siege in Gaza, highlighting the impact of technology blockades and technology warfare on crippling humanitarian operations, making it harder for populations to survive and for aid agencies to fulfill their mandates.

Building on the subject of technology in conflict-zones, David and Catalina discuss how digital tools and AI are becoming indispensable for humanitarian aid while at the same time they are being weaponized by non-humanitarian actors to facilitate the surveillance and persecution of civilians, testing humanitarian principles in new ways that we are only beginning to understand.

Catalina highlights an emerging "two-tiered system" within the humanitarian sector, characterized by a stark disparity in technological capacity. While large, well-funded organizations possess the resources to conduct rigorous research into uses of AI and deploy more sophisticated and secure tools, smaller grassroots NGOs are frequently left behind. This digital divide often forces smaller actors to rely on insecure platforms and tools, leaving their operations and the communities they serve vulnerable to cyberattacks and data breaches. Power imbalance like these also severely limit the negotiating leverage of smaller NGOs not only with global technology corporations but also with key stakeholders on the ground during a crisis.

The conversation concludes with a call for international political agreements to establish "red lines" that protect humanitarian data, much like the physical Red Cross flag protects aid workers under the Geneva Conventions. Looking forward, the guests explore the potential for defensive uses AI designed specifically to do things like safeguard human rights and provide medical guidance in conflict zones where traditional support has been severed. Finally, they call upon the academic and technology communities to deepen their engagement with the humanitarian sector, helping organizations of all sizes navigate the rapid evolution of AI and unlock its potential as a force for global good.

Interview notes: https://humanitarianaitoday.medium.com/humanitarian-technology-under-siege-a-conversation-with-catalina-rebollo-from-wired-en-espa%C3%B1ol-c7f143c98c97

Voices is a new mini-series from Humanitarian AI Today. In daily five-minute flashpods we pass the mic to humanitarian experts and technology pioneers, to hear about new projects, events, and perspectives on topics of importance to the humanitarian community.

In this flashpod, Radek Wierzbicki, CEO of Unsung Heroes, speaks with Humanitarian AI Today producer Brent Phillips about his team’s work connecting startups with humanitarian organizations and Unsung Heroes’ Humanity Badge initiative, a platform that builds the reputation of humanitarian and development workers and their organizations. They discuss Unsung Heroes’ core interests in advancing digital literacy and entrepreneurship and their work in Tanzania supported by the Polish Minister of Foreign Affairs and the Polish Embassy in Tanzania helping mentor young people interested in launching technology startups. They also discuss Unsung Heroes’ work distributing and training people on using computers and AI applications, and their work launching and partnering on technology incubators and accelerators advancing digital entrepreneurship.

Substack notes: https://humanitarianaitoday.substack.com/p/radek-wierzbicki-from-unsung-heroes

In this episode of the Humanitarian AI Today podcast, Siem Vaessen, CEO of Zimmerman and an IATI Governing Technical Board Member, and Sylvan Ridderinkhof, Data Engineer at Zimmerman, joined Brent Phillips to discuss the critical intersection of artificial intelligence, open data, and humanitarian collaboration amidst a rapidly changing and advancing landscape. Drawing on insights from the NetHope Summit, the guests highlighted a consensus that the sector must collaborate more closely around AI and open data sharing, a necessity largely driven by significant cuts in aid funding.

Siem, Sylvan and Brent discussed Zimmerman's long-standing commitment to the International Aid Transparency Initiative (IATI), an open data sharing framework widely used by humanitarian organizations to share granular information on aid activities, transactions and results. The discussion delved into Zimmerman's work and its future roadmap, focusing on enhancing the usability and quality of IATI data and on simplifying the complex process of reporting aid activities through IATI. They touched on the launch of Zimmerman’s updated AIDA (Aid Information Data Analytics) data platform and on other Zimmerman products and services tailored for the humanitarian aid and development communities and how they’re looking at ways of leveraging AI to improve search capabilities and support data enrichment processes. They also however caution listeners on risks posed by AI adoption, capable of potentially impacting IATI data quality and usability. Because AI models and agents aren’t natively trained to understand complex and subtle differences in ways that organizations report aid activities and publish their data, AI applications risk misinterpreting aid activity information. The use of AI applications to enhance and augment IATI data could add to these challenges, making complex, granular analysis of IATI data difficult or prone to misinterpretation without measures being taken to mitigate these risks .

Ultimately, the guests stressed that the progress of humanitarian technology hinges not just on powerful tools but on responsible innovation and a greater commitment to collaboration, including actively engaging with local actors and organizations that may traditionally be excluded from technical discussions on uses of AI.

Interview notes: https://humanitarianaitoday.medium.com/siem-vaessen-from-zimmerman-on-iati-collaboration-around-ai-and-the-development-aid-landscape-ebd36e0f20e9