Reflections from the KAIST–NYU AI Governance Summit

Ryanna

On February 6th and 7th 2026, I attended the KAIST-NYU AI Governance Summit as part of a di-Lab project. The summit was co-hosted by KAIST and NYU's Center for Bioethics. This meant that beyond the usual policymakers, legal experts and engineers we'd expect to find in the AI governance space, the room was also full of philosophers. This interdisciplinarity was special, because talking about interdisciplinarity and witnessing it are two different things… mainly because of the immediacy: a philosopher could respond to an engineer in the same breath, when usually (and to the public eye), these kinds of exchanges only happen across articles and scattered moments in time. The summit was precisely built for that immediacy: structured around working groups, aimed at actionable recommendations rather than a forum format. Watching so many disciplines take apart the idea of AI governance at the same time was disorienting, but also enriching in a way I'd rarely experienced.

Many things piqued my curiosity, and here are a few of them, picked across the different working group sessions, keynotes and panel discussion.

On framings : a keynote speaker from a closed session challenged the word frontier in "frontier AI." He argued that frontier is already a political choice, it encodes hierarchy and naturalizes competition. And the benchmarks we use to define it are themselves unstable. He warned that we might be racing toward something we cannot precisely define, measuring it with tools we cannot fully trust. Some framings we take for granted do a lot of work we are not aware of.

On regulation: current risk frameworks classify AI systems one by one, but systems that don't individually qualify as high-risk can still produce serious harm together. Neither the EU AI Act nor Korea's Basic Act accounts for this. Someone compared it to early environmental law which targeted individual pollution sources while missing the cumulative damage. Consequently, no one ends up accountable because no one caused the harm alone. This is only one issue among many, but it resonated with me more than the others as this specific case really conveyed a sense of danger, amplified by the fact I couldn’t quite see any viable solution to it so far.

On human agency : Dr. Kwame Anthony Appiah (NYU, Philosophy and Law) introduced constitutive deskilling: the idea that AI threatens the cognitive capacities that make us who we are : our discernment, our creative thinking. ‘It is especially bad not just because that kind of deference to others deprives us of an important moral autonomy, but because cumulatively, it deprives us of the capacity to manage and think about what is valuable in our own life.’ This was another moment where the danger felt tangible, but also personal.

On AI governance in Asia: a conversation that mattered to me particularly, as someone who came from the West to Asia to study AI governance. Rorry Daniels (Asia Society Policy Institute) noted that Western discussions tend to treat AI as a risk to manage, but much of Asia treats it as a ‘force multiplier’, a tool for immediate political and social problems. For instance, human-centric language in Asian policy isn't aspirational in the Western sense, but strategic, because social legitimacy is a precondition for adoption. What this made clear is that governance doesn't start from a neutral place, the assumptions about what AI is already shape what governing it means.

What I'll remember most was hearing people at the top of their respective fields in direct conversation with each other, each contributing their own piece to the puzzle. This is the kind of slow institutional work that AI governance demands when approached as more than a technical question. Being exposed to these exchanges early in an academic journey is a privilege, and not only for the formal learning that comes from the ideas being discussed. There's also the informal learning that comes from simply being in the room and hearing the process : what words are being chosen, what makes them hesitate. That kind of grounding is what I want to carry into my research, so that what ends up on the page stays connected to what happens in the room.