European Parliament Launches Its Own AI Platform for MEPs

The European Parliament has shown MEPs and officials the first demo of EPGenAI Hub, an internal artificial intelligence platform meant to replace the chaotic use of public chatbots for drafting speeches, amendments, and parliamentary questions. The system combines models running on the Parliament's own infrastructure with external services, including OpenAI's ChatGPT and Anthropic's Claude Sonnet.

What EPGenAI Hub Will Do

The platform will let users choose the model suited to the sensitivity of the task. Classified or politically sensitive material is meant to stay on models hosted inside the Parliament, such as Llama, GPT-OSS, or Mistral Small, while for less sensitive tasks MEPs and their assistants will be able to turn to more advanced external models, ChatGPT and Claude Sonnet, running outside the institution's infrastructure.

The goal is simple: today nothing stops a parliamentary assistant from pasting a fragment of a draft directive or a confidential memo into the free version of ChatGPT on their personal account. EPGenAI Hub is meant to offer an alternative under institutional control, rather than banning AI outright, which would be difficult to enforce anyway.

The Hallucination Problem in Amendments

The push to build the platform came from signals within parliamentary committees. Committee chairs noted a growing number of amendments requiring additional translation, repetitive editorial patterns typical of text generators, and references to provisions that simply do not exist.

Parliament has not been immune to AI slop and misleading AI-generated content - European Parliament official

In other words, unchecked use of public chatbots is already leaving traces in official legislative documents, including fabricated legal references that no one verified along the way.

Transparency and the AI Act

MEPs are currently not required to disclose whether a given text, amendment, or parliamentary question was produced with the help of AI. Brando Benifei, one of the architects of the EU's AI Act, says this must change as the regulation's next transparency rules come into force.

MEPs must lead by example as the AI Act's transparency rules become applicable - Brando Benifei, co-author of the AI Act

Not everyone in Parliament is enthusiastic. Irish MEP Barry Andrews of the Renew group has openly voiced concerns about automating tasks that MEPs were democratically elected to perform.

these are core democratic functions which voters elected us to perform - Barry Andrews, Renew Europe MEP

Conflict With Digital Sovereignty

The choice of ChatGPT and Claude Sonnet as external options raises questions about consistency with the EU's rhetoric on technological sovereignty. Brussels has spent months talking about reducing dependence on American digital infrastructure providers, while its own legislative institution plans to bring OpenAI's and Anthropic's tools into official use alongside the European Mistral.

For Polish readers, the most relevant point is the institutional parallel: the same dilemmas, data confidentiality versus the convenience of public chatbots, apply to every public administration body and large company in Poland. Government systems such as ExtrAct, and plans for a national registry of AI use in public offices, are moving in a similar direction to EPGenAI Hub, though so far without integrating commercial foreign models.

The September timeline is not yet final, the platform is still in testing, and its full scope of features may change before public launch. What will matter most is whether Parliament actually forces the use of a controlled environment instead of staff members' private accounts on public chatbots, since the tool's mere availability does not guarantee that.