We Can Steer This: How the European Parliament Told the Story of AI

ECPR Standing Group | 23 August 2026

European Parliament https://www.europarl.europa.eu/portal/en

Anne Goldmann

When the EU’s Artificial Intelligence Act was finally adopted in 2024, most of the public debate had already moved on from how the Act regulates to whether it should. Headlines focused on the moments when it looked as if everything might still fall apart: the dispute over facial recognition and law enforcement powers¹, the trilogue negotiations that nearly collapsed², the industry lobbying that shadowed the final months³. These flashpoints are usually what people remember about the road to the AI Act.

Behind these visible and widely discussed developments lies a more structural question: what kind of story did the European Parliament actually tell about AI while it was co-shaping this law? In our recent article in Government Information Quarterly “Narrating the AI Act: Policy narratives on artificial intelligence, their construction, and evolution within the European Parliament”⁴, we set out to answer exactly that.

AI narratives in the European Parliament

Most existing research on AI policy narratives has focused on strategy documents and white papers produced by executives such as national governments⁵ or the European Commission⁶. The Parliament’s own discursive role has received far less attention, even though it was a key actor in shaping the final text of the AI Act, most visibly in the fight over remote biometric identification and its use by police and law enforcement authorities.

For our study, we didn’t focus on the legal text of the AI Act itself; instead, we built a corpus of 18 documents produced by the European Parliament between 2019 and 2024: records of hearings by the special committee on Artificial Intelligence in the Digital Age (AIDA)⁷, working papers on topics from AI and health to AI and the labor market, committee reports, and the two plenary debates dedicated to the AI Act. Within these documents, we manually identified 500 narratives, small stories related to AI in which someone or something acts in a specific setting.

To analyze them systematically, we used the Narrative Policy Framework (NPF)⁸, a political science approach that breaks every policy narrative down into four building blocks: characters (who acts in which position – as a hero, a villain, a victim, or a beneficiary), a plot (a story of either rising, decline, control or helplessness), a setting (the context), and a moral (the proposed solution). The NPF lets us ask not just what MEPs said about AI, but how they structured what they said – the underlying grammar of the debate, so to speak.

What it tells us (and what it doesn’t)

It is worth being precise about what this kind of analysis reveals to us and where its limits lie. We studied the structure of the narratives told in Parliament, not the content of every individual argument or speech. So, what does this structure look like?

On characters, the picture is strikingly one-sided: nearly half of all character codings were heroes, and beneficiaries were the second-largest group. Villains and victims made up a much smaller share. In other words, when MEPs told stories about AI, they overwhelmingly cast someone, e.g. the EU, or something, e.g. AI, as the actor responsible for success rather than harm.

The plot tells a similar story: again, nearly half of all plot codings were stories of control, and another third were stories of rising. Stories of decline or helplessness were comparatively rare. Put simply: the Parliament’s underlying understanding was “we can steer this, and it will make things better”, not “this is happening to us”.

The setting and moral categories give us more content-specific details. Settings in which the topic of AI was placed ranged from security and law enforcement to global economic competition, with frequent references to existing European law and fundamental rights. The dominant moral, unsurprisingly, was the call for a risk-based regulation, followed by claims about protecting democracy and the need for not only a joint European but also an international approach.

Perhaps the most striking finding, though, is how stable all of this was over time. We expected the release of ChatGPT 3.5 in November 2022 to function as an external shock that would visibly shift the Parliament’s narrative components – and other research on parliamentary framing has indeed found such shifts⁹. But at the level of narrative structure, we found no such break: the plenary debates held well after ChatGPT’s release contained roughly the same share of stories of control and rising as the documents published before it. The surface-level framing of AI issues may have adapted; the deeper storyline, presenting AI as governable and beneficial, did not.

Why AI narrative research is more than “just another technical discourse”

Debates about new technologies and their consequences are as old as societies undergoing technological change themselves. It would be easy to file AI narrative research under this well-worn category and move on.

But there is something qualitatively different happening in the broader AI debate. Increasingly, AI and its potential capabilities are being positioned not merely as tools that governments and administrations can use, but as something that could substitute for political judgement itself, a better way of deciding than politics itself. And while efficiency gains in administration and bureaucracy are, of course, worth pursuing, a discourse that quietly displaces the political as such is a different matter entirely. Scholars such as Timnit Gebru have long warned about the costs of uncritically adopting techno-solutionist narratives around AI – narratives that ultimately tie technology’s promise to nothing less than the future of humanity itself¹⁰.

In the material we analyzed, the European Parliament’s narratives don’t yet go that far – MEPs consistently cast themselves and European law as the actors in control. But this framing is not equally dominant across the whole AI discourse. Outside Parliament, and especially among tech industry voices, the narrative increasingly leans toward AI as an alternative¹¹, even a superior one, to political deliberation¹². A shift worth watching closely, since narrative research shows that stories can travel across institutions.

This is, ultimately, why we think narrative analysis matters for AI governance research: laws are not only made through legal text and formal votes, but through the stories legislators tell about what a technology is, what risks it poses, and what kind of future it should enable. Those stories shape which problems get taken seriously and which get quietly sidelined. If you want to dig into the full findings, including the breakdown by document type, our GIQ article is available here⁴.

Anne Goldmann is a postdoctoral researcher at the Center for Advanced Internet Studies (CAIS), Bochum (Germany). Her research focuses on digital governance, artificial intelligence, and democratic innovations.