The European Parliament is concerned about how quickly deputies are using artificial intelligence and has developed a solution to regulate it. The institution plans to provide access to AI models developed by OpenAI, Meta, Anthropic and the French startup Mistral, offering deputies and their assistants a secure alternative to public services.
European deputies and their staff are increasingly using AI to write speeches, draft parliamentary inquiries and even amendments to laws. To regulate the use of this technology, the European Parliament is creating its own official internal platform, EPGenAI Hub. According to Politico, it is intended to be a safer alternative to public tools currently used in legislative work.
Implementation and risks to the democratic process
At present, the EPGenAI Hub is in the testing phase and, according to two anonymous sources, could be rolled out as early as September. The project responds to concerns from committee chairs about the rising volume of AI-generated documents. Officials noticed the use of the technology after a surge in amendments requiring translation, a distinctive writing style, and references to non-existent norms.
Brando Benifei, a deputy from the Socialists and Democrats group who led the development of the AI Act, said the parliament does not always avoid mistakes or misleading content created by AI. He added that MEPs must set an example in complying with transparency rules. Currently, about 2,100 people, or 20% of the institution’s staff, use AI daily.
Barry Andrews, chairman of the Development Committee, acknowledged that many deputies already use AI to prepare speeches, parliamentary inquiries and amendments. “This is their choice, and I understand it given our workload, but to me these are foundational democratic functions that voters have entrusted the parliament to perform,” he said.
Committee heads have asked Secretary-General Alessandro Chiocchetti to develop recommendations for AI use. Chiocchetti suggested the issue may fall under the competence of the Constitutional Affairs Committee, which would need to amend the internal rules to impose stricter guidelines. Today, deputies are not required to indicate whether their work was created with AI assistance.
Technical features and digital sovereignty issues
The beta version of the EPGenAI Hub lets users choose between models running on the parliament’s internal infrastructure and high-performance external services. Locally deployed models include Llama from Meta, GPT-OSS from OpenAI and Mistral Small. External services are represented by ChatGPT from OpenAI and Claude Sonnet from Anthropic.
The platform’s deployment broadens the institution’s openness to American companies, which contrasts with Brussels’ aim to strengthen digital sovereignty and reduce dependence on U.S. technology. Earlier, the Parliament had already replaced Google with the French search engine Qwant and disabled AI features on deputies’ tablets due to cybersecurity risks.
The parliament’s press service noted that the list of available models could change as new open or sovereign EU models from cloud providers sign contracts with the European Commission.
Source: Politico
