Artificial intelligence dominated discussions at this year’s World Economic Forum in Davos, from private roundtables on governance to high-profile panels on business transformation. Stanford HAI Co-Director James Landay and HAI Managing Director of Programs and External Engagement Vanessa Parli participated in the week-long summit, contributing to keynote speeches, panel discussions, and conversations at stakeholder meetings.
Throughout the week, they heard a consistent theme. The idea was that while enthusiasm for AI remains high, the tone has shifted from hype to effective real-world implementation. Leaders want tangible impact and clearer accountability. Geopolitics also loomed large, shaping discussions around “sovereign AI,” open ecosystems, and the risks of over-reliance on a single country or company.
Below, Landay and Parli share what they heard and highlighted in this conversation about Davos and the ideas that dominated the AI agenda.
What was the atmosphere surrounding AI at this year’s Davos conference?
James Landay: People are still optimistic, but they’re becoming more realistic. Compared to previous years, I felt that there were fewer people who wanted to experiment with everything. More and more people are saying, “AI is needed now to realize real revenue.” The hype is still there, but there’s more pressure to show what’s working.
Vanessa Parli: While we still heard a lot of excitement about what AI can do, we were also happy to hear industry leaders asking how they can make the business case for responsible AI. This is important at a time when public trust in AI is very low, especially in Western countries.
What “external to AI” factors shaped the conversation the most?
Lundy: President Trump’s presence and comments, especially around Greenland, have cast a little bit of a shadow. People asked seriously, “What’s going on?” It fed directly into the conversation about sovereign AI that was already being built and what happens when partners (or markets) can’t be trusted.
The term “sovereign AI” came up a lot. What did people mean by that?
Randy: It meant different things to different people. Broadly speaking, many countries want more control over their AI future, often in response to geopolitical uncertainty and the dominance of big tech companies.
What I tried to emphasize is to first define your goals. Our HAI policy team has analyzed this, and one helpful framework is that countries often pursue sovereignty to protect things like national security, economic security and prosperity, cultural values, and other national resilience goals. Countries can then choose where to focus their efforts on the “AI stack,” including compute (GPUs, data centers), data, models, applications, and people. Different countries place different emphasis on different groups depending on their goals.
Parli: HAI is currently working on research to define different angles of AI sovereignty and the benefits of each. This explains what AI sovereignty means, what it consists of, and how different approaches can benefit you. Generating research-based insights is essential for countries to make good decisions.
Do you agree with the “build your own model” version of Sovereign AI?
Landay: It’s not the only option. Many discussions assume that sovereignty means “we control everything, so we need to build our own models.” I insisted there was another way. It’s open source. This means building capabilities that are shared internationally so that no single company or country controls them.
Parli: HAI generally leans toward open ecosystems. Open data and open models help build transparency and trust in technology and accelerate innovation. For any country to reap the true benefits of technology, users must trust it.
Is HAI doing anything concrete in that direction?
Randy: Yes. We announced a memorandum of understanding with ETH Zurich and EPFL. EPFL is the first partner in a broad global effort to collaborate on open models and related work. We are also in discussions with other governments and research centers.
What were people thinking about AI and the changing nature of work?
Landay: The ROI of AI has been talked about a lot. We heard less about replacing workers and more about adding workers and changing work processes. People said, with AI, how would work processes and their design change, how would roles change, and what new products could we develop? Someone gave a great example. If AI makes lending decisions in 5 minutes instead of 5 days, how will that change your product and what will it enable for your customers?
In the past, we’ve heard the hype that if we don’t start working on AI soon, we’re going to be in trouble. This year, people are starting to realize that they have to do it, but if they don’t do it wisely, it won’t lead to real benefits.
Where are the workers in a pinch?
Lundy: In one of my talks, I heard a major business unit executive at a large technology company say that he was responsible for growing his portion of the business by about $40 billion over the next three to five years without increasing the number of employees. This means that we are not cutting jobs, but we are not hiring people. They use AI to improve everyone’s personal productivity. The main point is that while mass layoffs from these types of companies may be avoided, they will not create large numbers of new jobs in the future. If I were a new graduate, I think I would be anxious. But in the long term, I and others remain bullish on AI creating more new jobs.
A lot of people were talking about AI agents. What do you think?
Mr. Landay: There were two ways to be an “agent.” One is practical implementation within companies (which is already happening), and the other is a broader vision of many independent agents negotiating information and money over the open Internet. I’m more cautious about the latter, especially when personal or financial data is involved. Significant research and infrastructure is still needed before it can be widely trusted by people.
What did you focus on during your panel discussion?
Parli: We reminded people that while there are many opportunities in AI, they are not guaranteed and we need to think critically about how AI is designed and deployed. For AI to benefit everyone, we need many voices in the conversation.
Landay: If we want AI to be successful and beneficial to society, three things are necessary, but none are enough.
-
A design process that takes into account not only the user but also the community and society is what I think of as human-centered AI.
-
Ethics education and professional standards for those building these systems.
-
Regulation, Policy and Law. Some actors cut corners or cheat, and as in other industries, society needs mechanisms to respond.
All three are important. And even if you have all three, there will still be problems. You need realistic expectations and the ability to react if things don’t go well.
