The European Parliament is fighting "neural slop": how AI chaos in lawmaking gave birth to its own platform, EPGenAI Hub
The European Parliament has encountered an unexpected problem: MEPs and their assistants are massively using neural networks to write speeches, inquiries, and amendments, leading to the emergence of "neural slop"—low-quality, template-based content with hallucinations and references to non-existent laws. Instead of a total ban, the institution's leadership decided to take a different path—launching its own platform, EPGenAI Hub.
According to my data, about 2,100 people—nearly a fifth of the staff—resort to AI assistance daily. The situation escalated in July when an anomalous surge in amendments requiring translation was recorded during a meeting of committee chairs. The texts were characterized by a distinct dry style and contained references to fictitious regulatory acts. AI Act rapporteur Brando Benifei directly called this a "wave of neural slop and fakes." Irish MEP Barry Andrews, who chairs the development committee, admitted that colleagues write using AI due to the enormous workload, although he himself considers such tasks a key duty of a legislator.
European-style sovereignty
The platform has already been presented at workshops. The official launch is scheduled for September. The beta version offers a choice between local models (Llama from Meta, GPT-OSS from OpenAI, Mistral Small) and external services (ChatGPT, Claude Sonnet). However, a paradox lies here: by expanding the presence of American AI giants, the parliament is contradicting Brussels' course toward reducing technological dependence on the US. Previously, the institution replaced Google with the French search engine Qwant and disabled the built-in AI on MEPs' tablets for cybersecurity reasons.
The press service clarified that the list of models may change: sovereign European solutions from new cloud providers are planned to be added to the system.
Not just Brussels
Similar incidents have been recorded in Russia as well. Journalists suspected the use of ChatGPT in the preparation of draft law No. 1126815-8—a source=chatgpt.com tag was found in a footnote of the explanatory note. The State Duma explained that an employee used AI only for quickly finding the original source, and the data itself turned out to be correct. The committee emphasized that drafting legislation requires deep knowledge, and a neural network cannot yet handle such a task.
My expert conclusion: The European Parliament has chosen a pragmatic path—not to fight the technology, but to channel its use into a controlled direction. However, dependence on foreign models remains a vulnerability. Until European sovereign solutions appear in EPGenAI Hub, Brussels' "digital sovereignty" remains more of a declaration than a reality.