Humanitarian AI Today

Ethna Ghosh, India Partnerships Lead with Renaissance Philanthropy, discusses the nonprofit organization’s operational strategy, her work and India’s role as a global hub for AI experimentation with Brent Phillips, Humanitarian AI Today podcast producer.

Drawing on her deep roots in the Indian development sector, Ethna shares insights from her experience as a founding member of GivingPi, India's first family philanthropy network at Dasra, where she specialized in building strategic partnerships and advising families on high-impact giving. From this vantage point, she provides an insider’s perspective on India’s emergence as a science and technology pioneer and a premier real-world testing ground, where localized AI applications can be deployed across a diverse population of 1.4 billion people with immense cultural, geographic and linguistic diversity to address a myriad of social challenges.

Elaborating on her role as Renaissance Philanthropy’s India Partnerships Lead, Ethna outlines how the organization accelerates scientific, technological and innovation breakthroughs by using a thesis-driven, time-bound fund model connecting donors with domain experts to tackle high-impact, underfunded projects. By empowering field leaders to identify and eliminate systemic bottlenecks to enable scalable change, the organizations acts as force multiplier to achieve large-scale, transformative impact rather than just deploying capital. By supporting a diverse portfolio of deep-science innovations, the organization seeks to spark a modern "Renaissance movement" where converging technologies like AI and space tech for example transform the global landscape.

Offering insight into the keys to success in environments like India, Ethna emphasizes "bottom-up" approaches that prioritize deep collaboration between Samaj (society), Sarkar (government), and Bazar (market). Highlighting rural maternal health workers as a case study, Ethna points out that involving government entities is a non-negotiable in India for any initiative aiming to achieve true national scale and effectively engage the "last mile" of the population.

Looking toward the future, Ghosh envisions a shift toward agentic and physical AI that moves beyond simple text prompts to become a voice-enabled "buddy" for those with limited digital literacy. She advocates for the development of inclusive models that strip away linguistic and gender biases to provide safe, life-changing access to justice, healthcare, and government services. This vision ensures that technology serves as a bridge for equality, providing every citizen with a reliable companion to navigate a rapidly changing world. She concludes with a powerful call for radical collaboration among foundations and NGOs, urging stakeholders to move past silos and genuinely partner to eliminate the bottlenecks holding back progress.

Co-produce by Humanitarian AI Today, this third pilot episode of The Inference Layer podcast bridges the technical complexities of AI deployment with the reality of humanitarian operations and dives into the transition from static models to autonomous agentic systems. On behalf of the Humanitarian AI Today podcast, guest host Patrick Hassan, an AI policy lead with a background in disaster response, interviews Federico Pierucci, Scientific Director of the Icaro Lab, to explore how the inference layer is becoming a site of significant systemic risk. The discussion provides a unique look at inference-time failures such as alignment drift and steganographic coordination that emerge only when multiple agents interact in production environments.

For humanitarian actors, the episode raises concerns regarding operating in an era of assistance automated by layers of AI agents. The dialogue highlights how multi-agent chains used for beneficiary selection or resource allocation for example can degrade, develop invisible biases or be weaponized or politicized by parties to a conflict. Federico explains that these risks can be compounded by a lack of safety benchmarks for things like underrepresented languages and dialects, which can lead to unpredictable jailbreaks or administrative failures in the field.

The episode provides an inside look at pioneering research being carried out by the Icaro Lab, a Rome-based laboratory specialized in AI safety in collaboration with the Sapienza University. The lab focuses on mechanistic interpretability, a technical field dedicated to understanding the internal attention heads and decision-making units of an AI to decipher how it truly processes information. The discussion introduces the concept of Institutional AI, a proposed framework to manage these emerging xeno-behaviors through a governance graph. Rather than relying solely on prompt engineering or model-level alignment, Federico argues for a protocol-level solution that can manage misbehaving agents during inference. The episode is informative for professionals seeking to understand why AI safety must evolve from a localized technical challenge into a global institutional design problem, particularly in regions where traditional governance has broken down.

This particular episode moves beyond surface-level AI ethics and safety issues that the humanitarian community has been talking a lot about, to address inference-time vulnerabilities in agentic systems. This is an important topic because as the humanitarian community moves from developing and testing simple chatbots to incorporating autonomous multi-agent systems into humanitarian operations, we face new challenges that can have very serious consequences - making the 'inference layer' a new frontier for humanitarian risk.

Zineb Bhaby, AI Lead at the Norwegian Refugee Council, introduces NRC’s CLEAR (Crisis Learning, Early-warning, Anticipation, and Response) initiative and discusses the critical necessity of data collaboration in the humanitarian sector with Humanitarian AI Today producer Brent Phillips.

The CLEAR initiative is a three-year project supported by Twilio that is designed to build a digital "backbone" for humanitarian cooperation that the humanitarian community can collectively maintain and evolve. Zineb stresses that CLEAR’s goal is bring together humanitarian, academic and private sector partners through a consortium to integrate diverse data sources into unified early warning and early action systems, leveraging artificial intelligence and predictive analytics to transform how humanitarian organizations detect, prepare for and respond to crises.

Discussing CLEAR and challenges associated with the collection and use of data by aid organizations and the imperative to do better, Zineb nevertheless emphasizes that strict data governance remains a priority to protect the safety and sensitivity of information regarding vulnerable populations. By prioritizing an agile, safety preserving, open-source approach that bridges the gap between available information and field response, the initiative seeks to create a more resilient and unified technological foundation for the entire humanitarian ecosystem.

Voices is a new mini-series from Humanitarian AI Today. In short daily flashpods, Voices passes the mic to guests to learn about new projects, events and advances in artificial intelligence and discuss topics that are important to the humanitarian community.

In this flashpod, Lukas Borkowski, Senior Director of Strategic Partnerships at Viamo shares how artificial intelligence can serve the billions of people who remain offline and rely on basic mobile phones. In a conversation with Humanitarian AI Today producer, Brent Phillips, Lukas spotlights the reality that most people in lower income countries live their lives largely offline and disconnected from the benefits of emerging AI applications while at the same time live under mobile network coverage.

Lukas describes how Viamo works directly with mobile network operators to negotiate long-term partnerships that enable national-scale, toll‑free hotlines and behavior-change campaigns and he describes how Viamo is rapidly expanding voice-first gen‑AI experiences for use cases like rural health worker hotlines and disaster-preparedness campaigns.

He outlines Viamo’s cloud and in‑country server architecture, their use of generative AI and speech technology in local languages, and their specialization in behavior-change communication design that is tailored to specific geographies and demographics. Offering examples from public-health collaborations, he illustrates how voice-based generative AI can support and provide both community members and frontline workers with accessible information, advice and decision support.

Touching on broader ecosystem challenges, Lukas highlights the lack of high-quality speech technology for many African and Asian languages and calls for more investment, standardized tooling, and collaboration with aggregators like Viamo rather than fragmented pilots and one-off solutions. He calls for partners who bring clear behavioral objectives and a willingness to deploy imperfect but improving tools, arguing that waiting for perfect technology delays agency for people who urgently need trustworthy information. Looking ahead, he envisions seamless voice experiences where, in a single call, users can learn about services, ask personalized questions, and complete tasks.

Voices is a new mini-series from Humanitarian AI Today. In short daily flashpods, we pass the mic to guests to tell us about new projects, events and advances in artificial intelligence and to discuss topics that are important to the humanitarian community.

In this flashpod, Golestan (Sally) Radwan, Chief Digital Officer with the United Nations Environment Program and a Board Member of the Allen Turing Institute speaks with Panagiotis Moutis, Assistant Professor at the City College of New York and a member of the Climate Change AI initiative about AI’s environmental “sustainability paradox” with Humanitarian AI Today Producer, Brent Phillips.

Balancing technology’s potential to solve complex environmental problems against AI’s ecological costs, high energy consumption, water usage and e-waste, Sally and Panos emphasize that AI is not a magic solution but a complex equation where the new tools that we’ll use to save the environment are themselves taxing its resources, and suggest that AI’s value must be weighed against costs and resources that the technology draws away from other humanitarian and environmental needs.

The participants explore the potential of on-device machine learning to reduce the environmental footprint of AI by shifting workloads from data centers to local hardware and discuss the critical role of data infrastructure and global cooperation in addressing climate change. Sally touches on the challenges of data interoperability, noting that too many different standards exist for environmental data, which complicates the "multivariable analysis" needed for accurate climate forecasting. Panos offers a somber closing perspective, likening the struggle against climate change to a war where key battles may already be lost. He argues that AI's greatest potential might lie in creating clear, uncurated narratives to help the public and politicians grasp the existential urgency of the crisis. To help address this need for reliable information, Sally highlights the launch of EnvironmentGPT, a new tool designed to make environmental science easier to access and understand.

Humanitarian AI Today is a community-led initiative advised and co-produced by collaborating organizations and technology companies. Amidst a fragmented landscape, the podcast serves as a recognizable channel for organizations, donors, and innovators to collectively use to report AI projects, events and advances, turning raw insight into collective intelligence.