The European Parliament goes on the offensive: how Brussels bureaucrats decided to combat AI slop and "hallucinations" in lawmaking
The European Parliament, facing a genuine invasion of low-quality content generated by artificial intelligence (so-called "neural slop"), has adopted an unexpected but pragmatic solution. Instead of a total ban on AI tools, the institution's leadership is launching its own secure platform, the EPGenAI Hub. This move is not merely an attempt to regulate the use of neural networks, but a strategic response to the growing problem of "hallucinations" and outright fakes infiltrating legislative documents.
Scale of the Problem: One-Fifth of Staff "Hooked" on AI
Alarm in the corridors of EU power peaked after a July meeting of committee chairs. It emerged that around 2,100 parliamentary employees—nearly one in five—use neural networks daily to draft speeches, inquiries, and amendments. However, the issue lies not in the use itself, but in its consequences. Analysis revealed an abnormal surge in amendments written in dry, formulaic language and containing references to non-existent laws. This is a classic example of language model "hallucinations," where fabricated data is presented as fact.
MEP Barry Andrews, chair of the development committee, acknowledged that colleagues are overwhelmed and see AI as a survival tool. However, in his view, writing texts remains a key duty of a parliamentarian, and as long as legislators are not required to disclose their use of neural networks, the problem will worsen.
Own "Sovereignty" via Foreign Models: The Paradox of EPGenAI Hub
The new platform, whose beta version has already been presented to MEPs, offers a choice between local models (Llama from Meta, GPT-OSS from OpenAI, and Mistral Small) and external services (ChatGPT and Claude Sonnet). The official launch is scheduled for September. Yet here lies the main paradox: in trying to combat digital dependency, Brussels is effectively expanding the presence of American AI giants. This comes amid a push to reduce technological reliance on the U.S., under which the parliament has already replaced Google with the French search engine Qwant and disabled built-in AI on MEPs' tablets for cybersecurity reasons.
The parliament's press service clarified that the list of models will expand, and in the future, "sovereign" European solutions from new cloud providers are planned to be added to the system. But for now, this is more a gesture of desperation than a well-thought-out strategy.
Russian Connection: ChatGPT in the State Duma
Similar incidents are occurring in Russia. Journalists suspected the use of AI in drafting bill No. 1126815-8. An explanatory note was found to contain a footnote labeled source=chatgpt.com—a typical "trail" left when copying from a chat. The State Duma explained the situation by stating that a staff member had used the neural network only to quickly find the original source, and the data itself turned out to be correct. The committee emphasized that drafting bills requires deep knowledge, and a neural network is not yet capable of such a task.
Expert Opinion: The European Parliament's initiative is an attempt to "tame" a wild beast, not kill it. Creating its own platform is a right step, but it does not solve the fundamental problem: as long as MEPs are not held accountable for the quality of AI-generated content, "neural slop" and "hallucinations" will remain part of the legislative process. The market demands not just tools, but clear rules of the game and verification mechanisms.