MPs are using AI to help write parliamentary motions – without disclosing it

The two research teams have developed and used AI detectors trained on thousands of political motions in Sweden and the UK. What surprised the researchers was not that politicians were using AI – but that not a single one disclosed this openly.

“Regardless of the extent to which AI has been used, there may be risks associated with a lack of transparency. We also note that texts produced with the help of AI are often very well-written and persuasive, which increases the risk that errors will not even be spotted by the writers themselves. Furthermore, it is a matter of trust from the recipient’s perspective,” says Minerva Suvanto, the report’s lead author and a PhD student in the division of Vehicle Engineering and Autonomous Systems at the Department of Mechanical Engineering.

She refers, for example, to the SOM Institute’s report “Swedish AI Opinion 2023–2025”, which examines the public’s views on AI in various contexts. Among other things, it shows that, overall, 88 per cent of Swedes are quite or very concerned that “false information generated by AI may influence democratic elections”.

Calls for greater transparency

As the use of AI has become so widespread in society, it has recently become common practice to label texts or images produced with the aid of AI, including in the media. However, until very recently, there were no binding rules requiring such disclosure in Sweden. On 2nd August 2026, however, the European Commission’s new transparency rules came into force. The AI Regulation sets out how and when, for example, companies and public sector bodies must label their content to indicate whether AI has been used and, if so, to what extent.

“In our research, we have focused neither on regulatory frameworks nor on factual content, but have kept strictly to the language itself. However, it will be interesting to see how the new rules will be applied in Sweden and whether this will affect the drafting of parliamentary motions in the future. Although the AI Regulation is rather vague, and there are exemptions if AI is used solely for editing, politicians may reassess how they report their use of AI – given that they want to maintain voters’ trust,” says Suvanto.

To determine whether AI had been used to any extent in parliamentary motions, the researchers created bespoke AI detectors – one for Swedish texts and one for English. They began by analysing 1,400 Swedish and 1,400 British motions written between 2014 and 2020 – a period before established chatbots existed and when all motions were written by humans. To generate equivalent texts using AI, they first had a chatbot summarise each original text and, based on that summary, write a new text of its own. Using the large amount of training data – comprising both the original and the AI-generated versions of each text – the researchers were able to create language detectors that can determine with a very high degree of accuracy whether AI was used in the writing process.

The detectors report everything transparently

There are already a number of online tools that claim to be able to determine whether a text is AI-generated or not. What sets the method developed at Chalmers apart is that it openly shows its language analysis. Using colour coding in the text, one can follow the AI detector’s reasoning.

“We use what is known as interpretable AI. This means that we can not only investigate whether AI has been used, but also where it is evident in the text. But you can’t say that individual words are ‘AI words’ – it’s more complex than that,” says Mattias Wahde, Professor of Applied Artificial Intelligence at the Department of Mechanical Engineering.

To check how well the detection system worked, the researchers tested their method on a large number of parliamentary motions from 2021, when the current AI support was not available. The assessment was incorrect on only one occasion.

When all motions from 2022 up until this spring were analysed, the researchers noted a clear increase in texts where AI had been used in some form. Having been close to zero during the 2021–2022 parliamentary year, almost one in ten motions contained some AI-assisted text during the 2025–2026 parliamentary year.

The researchers found no significant differences between British and Swedish parliamentarians – nor between different political parties.