When Legislators Meet AI
Inside the Athens Democracy Forum’s two-day Parliamentary Caucus on technology, trust, and the future of representation
BY BEATRIZ REY
Last week at the Athens Democracy Forum, Members of Parliament from places such as Colombia, South Korea, the Netherlands, Wales, Portugal, Canada, and the Virgin Islands gathered to reflect on what makes them indispensable to democracy, even as they explored how artificial intelligence could transform their work.
In a two-day Parliamentary Caucus organized by the Better Politics Foundation, they participated in sessions under the theme “AI, Trust, and Leadership: Parliamentary Gathering for Democratic Resilience,” examining how technology might strengthen, rather than erode, the human core of democratic politics.
On the first day, legislators grappled with what it means to serve in an AI-shaped world, learning the basics of the technology, debating its ethical implications, and reflecting on how it might reshape trust between citizens and representatives. On the second, they rolled up their sleeves to design AI tools tailored to their daily work. One idea stood out: a personal “think tank,” an AI assistant that could help MPs generate ideas, analyze policy, and reclaim ownership over the knowledge processes often delegated to staff or consultants.
A special note from the gathering: South Korean MPs from opposing parties joined the same discussions – a small but powerful reminder that, even in polarized times, dialogue and imagination remain at the heart of politics.
Day 1: Understanding their role in a AI-shape parliament
The first day was dedicated to rethinking what it means to be a legislator in an era defined by artificial intelligence. Odanga Madung, a Technology and Human Rights Fellow at the Harvard Kennedy School, opened the session with an accessible introduction to the technology, reminding MPs that AI is more than just Large Language Models (LLMs) like ChatGPT. It also includes machine learning and deep learning systems that already influence daily governance and policymaking (on that note, the POPVOX Foundation team released an in-depth report on AI in parliaments, and is now working on the second edition).
Madung’s core message was clear: legislators must understand AI if they want to shape its future. “Technology follows power,” he said. “The laws you pass will determine its shape, its winners, and its losers.” He also urged MPs to reflect on how AI will transform their own work. It was agreed, for instance, that the traditional role of speechwriters may soon disappear. But Madung also warned that AI could upend trust in politics, as lawmakers become increasingly vulnerable to deepfakes and other forms of misinformation.
Trust was precisely the theme of the afternoon session, led by James Weinberg, a Senior Lecturer in Political Science at the University of Sheffield, who broke it down into three elements: competence, integrity, and benevolence. Whether AI can help rebuild these remains an open question.
The day closed on a reflective note. MPs asked themselves what their purpose is and what makes them indispensable within the political system. One participant summed it up: while AI can boost productivity, it cannot replicate empathy, creativity, or the human connection at the heart of representation.
Day 2: Imagining AI Tools for Parliaments
The second day began with a technical overview by Vassilis Katsouros, from the Institute for Language and Speech Processing, who explained how LLMs are built. Legislators were intrigued to learn, for instance, how words are broken down into “tokens” during the AI training process. After the session, they split into two groups to design an AI tool that could serve their everyday needs as Members of Parliament.
I joined one of the groups. It was revealing to see how uncertain many still are about what they actually want from AI. What began as a plan to create a tool that would “assist MPs on policy, like a global policy shop” gradually evolved into something closer to an “idea generator.” Soon, they were asking themselves fundamental questions: What problem are we trying to solve? Is it about drafting bills, engaging with constituents, writing speeches, or managing party conflicts?
At one point, I asked whether they really needed such a tool, given that they already have staffers, research services, and consultants to perform these tasks. Some noted that bias in staff work can be a concern. But for many, the motivation was about autonomy. As one member told me, “I know staff and consultants are using AI to do what I need. I want to be the one to do it. I want to write the prompt.” For him, the ideal solution would be a language model that functions as his own “think tank.”
James Weinberg, who was in my group, pushed the discussion into ethical territory. Should conversations with AI be deleted for safety reasons? Should the system always agree with the user or be trained to challenge them? How could it generate policy suggestions without falling into hallucination?
At the end of the session, both groups presented their prototypes. It was inspiring to watch legislators from different countries and political ideologies working together, not to defend positions, but to imagine new possibilities. As a researcher, I’ve been thinking deeply about how to reimagine politics, and I believe that the answer lies in spaces like this one – where ideas about democracy and technology meet in practice.
Modern Parliament (“ModParl”) is a newsletter from POPVOX Foundation that provides insights into the evolution of legislative institutions worldwide. Learn more and subscribe at modparl.substack.com.
