Taking Open Source Into the AI Era
re:publica/Gregor Fischer CC BY 2.0 Mark Surman is president of Mozilla, where he oversees a portfolio of organizations working to keep the internet open and accessible to all. After joining the organization in 2008, he spent over a decade building out its philanthropic arm. Since 2021, he has led the nonprofit’s transformation for the AI […]

re:publica/Gregor Fischer CC BY 2.0
Mark Surman is president of Mozilla, where he oversees a portfolio of organizations working to keep the internet open and accessible to all. After joining the organization in 2008, he spent over a decade building out its philanthropic arm. Since 2021, he has led the nonprofit’s transformation for the AI era, while maintaining its values of privacy and user empowerment. MIT Sloan Management Review spoke with Surman about managing a mission-driven technology organization through major transitions. This interview has been edited for length and clarity.
MIT Sloan Management Review: Mozilla is the corporate subsidiary of the nonprofit Mozilla Foundation. What are the challenges involved in leading an organization with such an unusual structure?
Mark Surman: Foundations that have a public purpose but also own companies aligned to that purpose are actually quite common in Europe. Lego, Ikea, Bosch — they're all owned by foundations. We closely studied Novo Nordisk, which is owned by a foundation focused on eliminating diabetes. The foundation makes large research grants while owning a company that produces diabetes treatments and generates revenue. So there are successful precedents, though it’s unusual in North America, and that comes with the challenge of not being easily understood by other companies or governments.
There are also internal challenges that come with our organization’s portfolio structure: We have tech companies, open-source projects, nonprofit activities, and venture investments. Even if the people who work at Mozilla are all committed to the same mission, they have different incentives and day-to-day priorities. They also come with varied career backgrounds and compensation expectations. We are a mostly distributed organization, too.
To better help keep everyone aligned and make the right trade-offs, we’re building a more intentional nonprofit holding-company function to improve coordination and transparency that includes board members who have cross-appointments to each of the company boards. We’ve also created a chief executive council comprising the heads of all our organizations. This helps us stay aligned and accountable while preserving the autonomy each entity needs.
You launched Mozilla.ai in January 2024. What does the company do, and what have you learned so far?
Surman: We launched Mozilla.ai because AI is today’s critical building block technology, just like the web was 25 years ago. What happens with AI now will determine future winners and possibilities. We have to play there in order to advance our mission.
The challenge here is holding to our values while adapting to changing user and market expectations, particularly around AI and data. Take Firefox: For years, we expressed our privacy values by collecting no user data. But today, people want adaptive, AI-powered features that require data to function. So we’re having to reimagine how to protect privacy while still delivering the experience users want.
So the Firefox team is already making progress on AI, but we needed an organization focused on it to draw some of the specialized talent we’re looking to draw. It started as a research project, but it’s now a startup looking to build commercial value around open-source AI.
How does Mozilla define responsible AI development?
Surman: There are two elements to trustworthy AI. The first is human agency: AI should be designed so that users have choice and transparency, helping them achieve their goals rather than manipulating them toward company objectives. Second is accountability: Systems must be transparent enough to diagnose problems and have clear frameworks for responsibility.
We’re also strong advocates for open-source AI. Open-source AI is essential for preventing centralized control and fostering genuine competition. Without it, we risk ending up with an AI universe that’s too concentrated and potentially harmful to society.