Should AI be used in the German parliament?

Hey Chat, write a speech on how AI limits critical thinking in political discourse to present in parliament. Sources? No need to double-check those.

Have you ever felt guilty about using AI to write your emails? Worry no longer; even those who are supposed to be leading Germany do. AI Chatbots have become the new speechwriters for many in parliament, but how can this be regulated in an environment that relies on human representation and critical assessment of societal issues?

Should the German parliament be permitted to use AI? 

Not for the first time, a representative in the parliament (Bundestag) has been revealed to have used AI Tools without disclosing them. The Federal Minister of Digital Affairs, Karsten Wildberger, uses AI to write his articles. Mario Voigt, the Minister of Thüringen, generated an entire speech with AI, quoting fake and hallucinated experts.

Now, Julia Klöckner, President of the German parliament, says AI use is not a “question of whether, but a question of how”. She refers back to the core values of parliament, saying AI has to be used responsibly, and even encouraged her colleagues to utilise AI at work. What this responsibility means on paper will be negotiated with the parliament's staff council, as stated by Tagesschau.

Florian Druckenthaner, speaker of the Federal Ministry of Digital Affairs, recognised that AI would never be able to replace human judgement in politics, but should rather be used as a “tool in a human process”.

The federal government has already made guidelines for the use of AI in federal management in the spring of 2025. Accordingly, AI has to be used responsibly and respectfully. This entails disclosing the use of AI and ensuring that the information gathered is accurate.

Government agencies have to follow additional guidelines regarding which systems are permitted and whether those systems comply with data protection laws.

Peter Sprong, the president of the Association for Speechwriters of the German Language (VRdS), emphasises AI-Literacy as the solution. Rather than using tools to generate texts, he suggests AI be used as a conversation partner to improve low-quality texts and overall reduce workload in his industry, while being used mindfully. According to him, the responsibility to critically assess the usage of AI lies with the writer.

The issue with AI-generated texts

So what is the problem? Why not just use AI? The issue with AI-generated writing, as well as using AI as a research tool to find sources, lies in the technology's false objectivity. Platform design makes it easy to think that these tools are all-knowing and free from human flaws and prejudices. Meanwhile, media research has shown that AI can exhibit discriminatory judgment. One example is a study on how different dialects lead AI tools to make racist and biased decisions.

Furthermore, AIs are trained by their developers; depending on the data they are trained on, they react in certain ways, creating much narrower products than assumed. 

Additionally, most AIs are based on algorithms. Algorithms are never objective; they do not provide customers with general information but with information that is frequently interacted with, regardless of the data's quality, thereby reinforcing users' stereotypes and prejudices. 

Finally, large language models usually work with probability, “guessing” what word is likely to come next. Especially in the early stages of AI, this led to the generation of what is called a hallucination: false text containing misleading information presented as fact. Nowadays, AI writing tools have even started to reference older hallucinated texts, closing the circle of misinformation. In parliament, all this is then combined with political power and real impact.