Sciences Po Hosts Two-Day Conference on AI, Democracy, and the Public Sphere
PARIS — Sciences Po has concluded a two-day event dedicated to one of the defining political questions of our age: how artificial intelligence is changing democracy and the public sphere. Held at the school’s 1 St-Thomas campus, the gathering brought together scholars, students, policymakers, journalists, technologists, and civil-society voices in a series of panels, keynotes, and workshops that sought to move beyond both techno-utopianism and doomsday narratives. The event, organized under the title “AI and the Public Sphere,” arrived at a moment when generative AI tools have become cheap, accessible, and widely embedded in everyday communication. From synthetic images and automated news feeds to algorithmic content moderation and microtargeted political advertising, the machinery of public conversation is increasingly controlled by machines as much as by humans. The central aim of the conference, according to the programme, was to ask what still remains public about that conversation, and what democracies need to do to protect it. The location itself was symbolic: Sciences Po is one of Europe’s foremost institutions for the study of politics, law, and social sciences, and its campus is a gathering place for students and researchers who are concerned with the future of civic life. With students gathered outside the entrance between sessions, the event had the energy of a university locked in urgent conversation with the world outside its walls.
The opening sessions framed the challenge in stark terms. Artificial intelligence is no longer a speculative future technology; it is operating now inside the systems that produce, distribute, and interpret information. Algorithms determine which posts appear in news feeds, which news articles are recommended, which political advertisements are shown to which voters, and which content is removed as harmful or misleading. The rise of large language models has added another layer of complexity, because these systems can generate plausible news stories, speeches, images, and audio recordings in seconds, making it far more difficult for ordinary citizens to know whether a statement, a photograph, or a video is authentic. Researchers at the event presented evidence on how synthetic content can amplify social divisions, intensify polarisation, and undermine trust in institutions. But the tone was not one of simple alarm. Several speakers noted that AI is also a potential tool for civic empowerment: it can be used to monitor government decisions, translate public documents into simpler language, help citizens navigate administrative services, and even facilitate large-scale deliberation. The central challenge, repeated throughout the day, is therefore not whether AI will affect democracy, but how, in whose interest, and with what safeguards. The idea of the public sphere, a concept familiar from political theory, was revisited and updated for the age of platforms, recommender systems, and synthetic media. What used to be a space of shared conversation in newspapers, town halls, and public squares is now a fragmented landscape of private platforms and algorithmically curated feeds. The question that hung over the entire first day was whether that landscape can still function as a space where citizens encounter one another, test arguments, and form judgements together.
Day one of the event focused specifically on the public sphere and the health of democratic communication. Panels explored the changing economics of journalism, the pressure on newsrooms to adapt to artificial intelligence, and the growing importance of fact-checking and media literacy. One session looked in depth at platform governance, asking how tech companies should be held accountable for the public consequences of their design choices. There was broad agreement that voluntary self-regulation has not succeeded and that governments must act, but the question of exactly how to act produced lively disagreement. Some speakers argued for stricter transparency requirements, so that researchers can examine how ranking algorithms work and how political content is recommended. Others called for new rights for citizens, including the right to understand when they are interacting with AI and the right to challenge automated decisions. A particularly well-attended workshop examined deepfakes and synthetic media, with demonstrations of detection tools and discussions of the limits of forensic approaches. The consensus was that no purely technical fix will be sufficient; citizens need educational resources, journalists need verification protocols, and the public sphere needs stronger collective infrastructure. Throughout the sessions, Sciences Po students played an active role, asking pointed questions and connecting the discussion to their own studies in political science, law, sociology, and digital humanities. The event was deliberately structured to be accessible, with several sessions open to the public and streamed online. By the end of the first day, a clear theme had emerged: the public sphere cannot be rebuilt simply by tweaking algorithms or enacting one law. It requires a broader democratic effort to define what kind of information environment we want, what role private companies should play, and what responsibilities public institutions must carry.
The second day of the conference turned more explicitly to democracy, elections, and institutional design. A morning keynote addressed the use of artificial intelligence in political campaigns, from automated voter outreach to microtargeting and the production of personalised content. The potential for abuse is obvious: AI can generate fake news tailored to the fears and biases of specific groups, flood comment sections with coordinated messages, or create synthetic recordings of candidates saying things they never said. Yet the speakers also reminded the audience that AI has already been used to support democratic participation in many countries. Text-generation tools can help citizens draft submissions to public consultations; machine translation can make political debates understandable across language communities; and predictive models can help governments respond to the needs of marginalised populations. The challenge, as one panellist put it, is to direct the technology toward democratic purposes rather than allowing it to be captured by the few. A significant part of the second day was devoted to the European regulatory response, in particular the European Union’s Artificial Intelligence Act. Several speakers described the Act as a historic milestone but warned that legislation alone will not solve the deeper problems. Rules need to be enforced, and enforcement requires resources, expertise, and political will. The role of independent courts and civil-society organisations was repeatedly highlighted as indispensable. Workshops in the afternoon explored new forms of civic technology, such as participatory budgeting platforms, AI-assisted deliberation, and tools that allow ordinary citizens to understand complex policy documents. There were also sessions on surveillance technologies, biometric identification, and the rights of individuals in the face of automated government decisions. A recurrent concern was the risk of exclusion: if AI systems are designed by and for wealthy, educated, urban populations, they may deepen existing inequalities. Speakers from the Global South stressed the importance of local voices and the need to avoid a model where a few large technology companies define the future for the rest of the world. Sciences Po’s interdisciplinary structure, with its multiple research centres and professional schools, provided a natural setting for these discussions, and the participants repeatedly crossed the usual boundaries between law, computer science, journalism, and political theory.
The event was not only an opportunity for reflection; it also offered a preview of Sciences Po’s academic programmes and its ongoing commitment to preparing a new generation for the challenges posed by artificial intelligence. Alongside the conference, the school announced details of two upcoming virtual open house events, both to be held in 2026, that are designed to welcome prospective students from around the world. The Virtual Undergraduate Open House day will take place on Saturday, 28 November 2026, and will allow participants to discover Sciences Po’s bachelor’s programme, its dual-degree options, and the first-year admissions process. The Virtual Graduate Open House day will be held on Saturday, 17 October 2026, and will showcase more than 30 master’s programmes offered across the school’s eight professional schools. For young people interested in the intersection of technology and public affairs, these online events are likely to be especially relevant. The two-day conference demonstrated that Sciences Po is treating artificial intelligence not merely as a technical topic for computer scientists, but as a core subject for political education. In the bachelor’s curriculum, students are encouraged to think critically about the social and political dimensions of technology; in the graduate schools, courses on data ethics, digital law, media studies, and public policy prepare students to become designers, regulators, and interpreters of AI systems. The virtual open house days will therefore serve as an accessible entry point for anyone who wants to understand how a social-science education can be used to shape the digital revolution. They also reflect a broader trend in higher education: no field of study is immune to the effects of artificial intelligence, and the most valuable education will be the one that teaches students to ask the right questions, not just to use the latest tools.
As the two-day event drew to a close, a consensus seemed to emerge from the many sessions and informal conversations: the conversation about AI and democracy has only just begun, and it must become broader, more inclusive, and more sustained. The speakers returned repeatedly to the need for new institutions, new professional norms, and investment in public-interest technologies. A public AI model funded by governments, transparent algorithms, and open data were among the ideas discussed. But the deeper message was even more important: democracy is not a machine that runs by itself. It depends on citizens who are prepared to listen, argue, and vote; on journalists who are willing to investigate; on civil servants who can act with integrity; and on universities that provide a space for critical thought. Sciences Po’s event at the 1 St-Thomas campus was a small but meaningful contribution to that larger democratic project. It was a reminder that the future of artificial intelligence will not be decided only by engineers and investors in faraway head offices. It will also be decided in classrooms, parliaments, courts, newsrooms, and public spaces where people come together to debate the kind of society they want. The two-day conversation is over, but the work it pointed toward is only beginning. With its forthcoming open house days and its commitment to public engagement, Sciences Po appears ready to remain at the centre of that work for many years to come.
