Brussels discovers AI can write summaries faster. That's the problem.
The European Union is turning to artificial intelligence to manage information overload-one of its oldest institutional challenges. Large language models now digest parliamentary proceedings, legislative files, and policy debates in seconds, a task that once consumed hours of human effort.
For journalists buried under documents and parliamentary staff sprinting between committee meetings, the technology offers genuine relief. A city built on summaries has found what looks like a solution: machines that write them faster and promise clarity.
But researchers at University College Dublin found something troubling buried in those efficiencies. When political speech passes through ChatGPT and similar systems, not every voice survives equally.
What the research shows
The study examined AI-generated summaries of actual European Parliament debates. Researchers fed transcripts into different large language models and measured whether the summaries accurately represented the distribution of political content across the Parliament's political groups.
They found "consistent positional and partisan biases" in how speeches were represented. Certain speakers were more likely to disappear from summaries, lose attribution, or receive less balanced representation once debates were condensed by AI systems.
"Bias shows up as systematic distortion," said James Cross, director of the Connected Politics Lab at University College Dublin, "with certain political positions being compressed, omitted or amplified relative to what MEPs actually said."
The study does not suggest AI systems are secretly ideological. The findings are subtler. The systems appear to compress certain political positions-particularly those from right-leaning groups like the European Conservatives and Reformists-while amplifying centrist ones.
Smaller political groups and speakers using less widely spoken EU languages face even greater risk of underrepresentation.
Why this matters in Brussels
Few people inside the EU institutions consume parliamentary proceedings in full. Most rely on extracts, speaking notes, and reports produced by others. Politics already reaches most people filtered through several layers of interpretation.
AI may simply become the next layer. But in Brussels, small omissions carry outsized weight.
A quotation dropped from a summary may vanish from a journalist's morning briefing. A flattened political argument may shape how policy professionals understand a legislative fight. A speaker omitted from an automated recap may disappear from the wider conversation altogether.
Over time, the abbreviated version of politics can become more influential than the original itself.
The timing problem
The EU has spent years positioning itself as the global capital of "trustworthy AI" through legislation like the AI Act. Meanwhile, AI tools are quietly creeping into daily workflows at the heart of European institutions.
On paper, it works. Brussels runs on information overload. AI offers a way to survive it.
But once AI-generated summaries become part of the infrastructure through which political debate is consumed, they stop being simple productivity tools. When political understanding increasingly depends on what these systems preserve, flatten, and omit, the design choices begin to matter politically.
Brussels has always depended on summaries to function. The question is what happens when those writing them are no longer entirely human.
Your membership also unlocks:
