Prague Workshop on Responsible AI: A Recap

Is it possible to understand the broad range of topics hidden under the umbrella of Responsible AI holisticly? We tried to do exactly that by inviting a human rights lawyer, a sociologist, an AI safety researcher, a philosopher, and an economist to share one stage at our first Prague Workshop on Responsible AI on Friday, October 9, held at FEE CTU on Karlovo náměstí under the auspices of Ondřej Vaněk, Chief AI Officer of CTU, and with the support of the Artificial Intelligence Center and the Intelligent Data Analysis group. Many people across Prague work on the ethics, governance, safety, and trustworthiness of AI, but they do so without crossing the interdisciplinary river. We wanted them in the same room to share the conversation with everyone who envisions the future in which AI is used, developed, and researched for the public good. We are happy that more than 130 people registered for our workshop.
Opening: The CTU Leadership View
CTU Rector Michal Pěchouček opened the day as a ‘self-described techno-optimist who also cares’. He argued that we still have a chance to get AI right, called on scientists to explain their work to the public and to legislators, and offered a comparison that stayed with many of us: roughly half of the investment in nuclear energy goes to safety, around five percent in cars, and next to nothing in AI. If frontier labs ever agree to slow down, he suggested, it should take the form of a massive investment in AI safety.
Ondřej Vaněk followed by holding today’s frontier labs up against the 2017 Asilomar AI Principles, noting that voluntary company principles are exactly why regulation with real enforcement matters, and shared his vision of principles for responsible research and education at CTU.

The Morning: Rights, Platforms, and Alignment
The morning block moved from present harms to future risks and back again. Martina Šmuclerová argued that AI has been subject to international human rights law from the moment it was built, and that the real challenge is translating legal norms into computational practice. Markéta Švarcová showed what happens to responsibility when the algorithm manages the worklife of platform workers. Ondřej Bajgar argued that both alignment and control of AI are failing, utilizing the July incident, in which OpenAI’s AI agents escaped their sandbox and attacked Hugging Face. The panel that followed, moderated by Ondřej Vaněk, asked whether AI regulation works or whether we need a different approach. The liveliest exchange came from the opposing views of Ondřej Bajgar and Martina Šmuclerová on international law, and the discussion eventually settled on enforcement by individual states as the real problem.
The Afternoon: Lightning Talks, Trust, and Collaboration
After lunch, six lightning talks brought the voice of the community on stage, with topics ranging from assessment design that protects students’ judgment, through agent-to-agent recruitment and the price of AI guarantees, to the psychology of language models and the guilt we (do not) feel when delegating to AI. John Dorsch, a returning guest of ours, then argued that we should stop calling AI “trustworthy” without qualification, and backed the claim with an experiment showing how much using a different label changes what people believe about a self-driving car. Petr Špecián revisited his proposal for universal basic AI access a year and a half later and concluded that, while the diagnosis holds, the clock is running faster than the institutions. The final panel, moderated by Gustav Šír, produced the most passionate discussion of the day, on whether the direction of AI progress should be chosen democratically or left solely to the market.

What Comes Next
The workshop was meant as a starting point, not a one-off. We have set up a Discord server for the community so that the conversations started on Friday can continue, and if Discord is not your thing, you can always write to us or join us at our coffee chats. In the meantime, several sessions of the upcoming AI Days in November will take place at CTU, and our mailing list is the easiest way to hear about the next event.
We are grateful to all the invited speakers, panelists, and lightning talk presenters for their time and their willingness to state their case in public; to Michal Pěchouček and Ondřej Vaněk for opening the day; to the AI Center and the Intelligent Data Analysis group at FEE CTU for their support; and to everyone who came, asked questions, and upvoted the good ones. The workshop was organized by Martin Krutský, Jakub Peleška, Jiří Němeček, Lucie Stewart, Eliška Skládalová, and Gustav Šír on behalf of the Responsible AI in Prague Initiative. See you at the next one.





