Watch the fireside chat recording.

 

Key Takeaways

• AI should strengthen care, not replace people. The promise of AI is not removing humans from the loop, but giving community health workers, clinicians, and local leaders more time and support for the relationships at the center of care.
• Friction can protect accountability and trust. AI should not become so seamless that users lose the chance to understand, question, and make informed judgments about how decisions are made.
• Ownership matters as much as innovation. Open-source approaches, local ownership, and country control can help ensure technologies reflect local priorities rather than the incentives and values of distant platforms.
• Global health has more leverage than it thinks. Nonprofits, ministries of health, and implementers can — and should — help shape the terms under which AI is deployed, using their influence to demand openness, accountability, and support for the infrastructure they depend on.
• Open-source security is an equity issue. Many global health organizations rely on open-source systems, but the people maintaining them often lack the tools and resources large companies have to keep their systems secure.

“The power actually rests in this room and rests in our organizations…you should be dictating the terms, not the other way around.” — Raffi Krikorian, Mozilla

At a moment when global health organizations are racing to understand how artificial intelligence (AI) will reshape their work, keynote speaker Raffi Krikorian, Mozilla’s Chief Technology Officer, delivered a different message at the Bay Area Global Health Alliance Annual Meeting: the most important questions are not about what AI can do, but who controls it, who benefits from it, and whether humans remain firmly in charge.

In a fireside conversation moderated by Alliance board member and Stanford Biodesign Director of Innovation to Impact Krista Donaldson, the discussion ranged from trust and accountability to community health workers, local ownership, and the future of open-source technology. Throughout the conversation, Donaldson pushed Krikorian to connect broad debates about AI to the practical realities facing global health organizations.

Drawing on experiences spanning Mozilla, open-source software, philanthropy, and his role as Board Chair of Alliance member Medic, Krikorian argued that AI’s future will be determined less by technical breakthroughs than by governance, ownership, incentives, and trust.

“We have to really ask ourselves, especially in this world, where does your thinking stop and where does the chatbot thinking start?” he said.

Krikorian opened with a story about a crash involving his Tesla operating in self-driving mode, using it to illustrate what he called the “moral crumple zone” — the tendency for responsibility to fall back on humans when AI systems fail. 

“The phrase that I like to use is, ‘I’m the moral crumple zone for the car,’” he said. “This system wasn’t able to decide what to do, so at the last minute it just threw it to the human.”

That example became a broader metaphor for AI deployment. Donaldson noted that many Alliance members work with community health workers and understand that technology adoption depends not only on what a tool can do, but whether users have reason to trust it. Krikorian agreed, warning that many AI systems are designed to project confidence, often beyond what users can reasonably evaluate.

“The bad version of that outcome is to go with the model that sounds the most confident,” he said. “It doesn’t necessarily have to be right. It just sounded the most confident.”

Yet his message was not one of caution alone. Krikorian repeatedly emphasized that nonprofits, ministries of health, and frontline implementers have more influence over the future of AI than they often realize.

“The power actually rests in this room and rests in our organizations,” he said. “You have more power in that conversation than you think you do.”

Rather than viewing themselves as passive recipients of AI tools, he argued that global health organizations should help shape how those tools are designed, deployed, and governed.

That conviction underpins Medic’s own AI strategy. When Donaldson asked what organizations building AI-enabled tools should do differently, Krikorian drew a distinction between automation and augmentation — a theme that surfaced repeatedly throughout the conversation.

“There was one lab working on automation, basically removing humans from the loop, and another lab working on augmentation — giving humans superpowers. I’m much more in that [superpowers] camp.”

Rather than replacing community health workers, he argued, AI should help them do their jobs better, freeing up more time for what humans do best: providing care, building relationships, and exercising judgment.

Donaldson also challenged him on the growing push toward frictionless AI systems. Krikorian’s response was one of the session’s most memorable ideas: some friction is not a flaw but a feature.

“There is benefit to friction,” he argued. “There is benefit to actually having a stumbling point.”

Rather than creating systems that operate invisibly in the background, he suggested that human-centered design should help users understand how decisions are being made and retain agency over those decisions. In his view, productive friction helps protect accountability, trust, and good judgment.

A recurring theme throughout the conversation was ownership. Krikorian advocated strongly for open-source approaches, local control, and country ownership of technology systems, arguing that who owns and governs technology may ultimately matter as much as the technology itself.

He described Medic’s ongoing effort to establish more locally governed organizations in countries where it works, arguing that technologies should adapt to local workflows rather than requiring communities to adapt to systems designed thousands of miles away.

“I firmly believe that the needs of Uganda and the needs of Kenya are similar but not the same,” he said.

Another theme was the vulnerability of the open-source infrastructure that underpins much of global health and the internet. Krikorian warned that increasingly powerful AI systems are becoming capable of identifying software vulnerabilities at scale, creating new risks for under-resourced open-source projects.

“Critical infrastructure is being run this way,” he noted, describing open-source digital infrastructure that in some cases is maintained by only a handful of individuals. “We need to understand who does critical infrastructure in this world. It’s volunteers who are the essential workers.”

He argued that global health organizations should not only understand the open-source systems they depend on but also help strengthen and sustain them. In his view, access to secure digital infrastructure is increasingly an equity issue: the organizations serving vulnerable populations often rely on tools maintained by small, under-resourced teams that lack the support available to major technology companies.

Beyond infrastructure, Krikorian warned that AI systems inevitably reflect the assumptions and incentives of their creators.

“If we don’t own every layer of this — or at least influence every layer of this — these systems might be doing things that don’t align with what we’re saying.”

When asked how organizations should prepare for the pace of change, Krikorian urged them to focus on AI readiness rather than rushing toward deployment: better data stewardship, stronger documentation, organizational learning, and a willingness to experiment.

“We can’t just ostrich ourselves at this moment,” he said. “Things are moving too fast.”

He encouraged organizations to share practical lessons, workflows, successes, and failures with one another, suggesting that peer learning may be one of the most valuable ways the sector can navigate rapid technological change.

Despite the challenges, Krikorian remained notably optimistic. When asked directly by an audience member whether society is “screwed,” he rejected the premise. “I’m actually a very optimistic person,” he said. His optimism rests on a belief that communities still have agency to shape how AI develops and is deployed. But that outcome is not guaranteed.

“If we all just sit around and do nothing and assume that it’s all inevitable, then it will be inevitable,” he said. “We actually have to do something.”

The conversation closed on a hopeful note. For organizations facing funding constraints, workforce pressures, and growing demands for impact, Krikorian believes AI can become a powerful tool — not by replacing people, but by amplifying their capabilities.

“I actually do think that this is a path, if done correctly, for us to still have the kind of impact we want on [using] potentially less dollars,” he said. “The goal is to augment the work we’re doing, not replace it.”

For a global health community grappling with uncertainty on multiple fronts, the message was both a challenge and an invitation: stay curious, stay engaged, and help shape the future rather than simply reacting to it.

 

See more on the Alliance’s 2026 Annual Meeting. Learn more about the Alliance’s work on AI and global health.

 

Speaker Bios

Raffi Krikorian, CTO, Mozilla; Board Chair at Mozilla.ai and Medic

Raffi Krikorian is the Chief Technology Officer at Mozilla. Krikorian has served on the Mozilla Foundation Board of Directors since 2023, on Mozilla.ai’s Board since 2024, and on the Mozilla.org Board since its inception. He also serves as the Board Chairperson at Medic. Krikorian brings a record of impact across sectors to Mozilla, with roles spanning tech, politics, media and philanthropy. He joins Mozilla from having been the CTO at Emerson Collective, where he led efforts to bring technologists into sectors like education, the environment, immigration, and economic mobility — and to help people in those sectors see themselves as technologists. He also hosted the “Technically Optimistic” podcast and Substack, exploring technology’s impact on society. Prior to that, he was the first CTO of the US’s Democratic National Committee, where he used data, technology, and digital security to support the election processes of Democratic candidates up and down the ballot; Director of Uber’s Advanced Technologies Center, where he led the development and rollout of the first passenger-carrying self-driving car fleet; and Vice President of Platform Engineering at Twitter, where he managed and built Twitter’s global infrastructure. 

Krista Donaldson, Director of Innovation to Impact, Stanford Biodesign; Alliance board member

As Director of Innovation to Impact at Stanford Mussallem Center for Biodesign, Krista Donaldson’s work focuses on ensuring that design tools and processes are broadly applicable across global markets. She is also part of the team establishing the East Africa Biodesign Program, which kicked off in early 2023. Stanford’s Biodesign program advances health outcomes and equity through innovation education, translation, and policy. As the former CEO of Equalize Health (formerly D-Rev), Donaldson led the design and scaling of disruptive medical devices to address global health inequities. To date, nearly 1M people – mostly children and young people – have been treated by one of Equalize Health’s products in 70+ countries. Peter Singer of the Effective Altruism movement called Equalize Health “one of the world’s best charities” because of its cost effectiveness and exemplary end-to-end processes. Donaldson has been recognized as a World Economic Forum Technology Pioneer, TED speaker, and a GLG Social Impact Fellow. She was also named one of Fast Company’s “50 Designers Shaping the Future.” Prior to Equalize Health, she was an Economic Officer at the U.S. Department of State where she managed part of Iraq’s reconstruction portfolio. She also worked at KickStart International (Kenya), and the design firm IDEO (USA). Donaldson holds a master’s degree in Product Design and a Ph.D. in Mechanical Engineering from Stanford University.