The European Parliament (EP) is attempting to launch its own internal artificial intelligence (AI) system. Known as the EPGenAI Hub, it aims to assist (and regulate) lawmakers and parliamentary assistants who increasingly rely on artificial intelligence to draft speeches, questions, and amendments,
The new platform will provide secure, in-house access to major models from OpenAI, Meta, Anthropic, and France’s Mistral, in a move designed to offer a safer alternative to public platforms and prevent sensitive legislative information from being leaked into commercial, uncontrolled databases.
The development of the hub highlights a growing dilemma within Brussels, where approximately 20% of all staff—around 2,100 people—currently use AI daily. Committee chairs have voiced rising alarm over the spread of ‘AI slop’ and hallucinated legal references appearing in official drafts. Brando Benifei, the lawmaker who led the EP’s work on the landmark AI Act, warned that MEPs must lead by example under the new transparency rules. Despite these concerns, there is currently no formal obligation for lawmakers to disclose if their legislative questions or amendments were generated by an automated model.
The rollout of the EPGenAI Hub also contradicts the EP’s recent push for digital sovereignty and its desire to reduce reliance on U.S. technology. Despite efforts to replace American services with European alternatives, such as by disabling built-in tablet AI tools and adopting the French search engine Qwant, the new hub relies heavily on American giants such as Meta and Microsoft-backed OpenAI. While the EP’s press service noted that the available models may change in the future to favor European sovereign cloud providers, the immediate implementation embeds American technology deeper into the core of the European legislative process.
