EU Parliament Builds Its Own AI Platform for Lawmakers

The European Parliament — the institution that shepherded the world’s first comprehensive AI law — is now confronting the technology inside its own walls. It is building an internal generative AI platform, called EPGenAI Hub, to give members and staff sanctioned access to models from Meta, OpenAI, Anthropic and France’s Mistral, according to Politico, which reviewed a beta version and spoke with people familiar with the plans. The platform is in testing and could begin rolling out as early as September 2026.

The aim is oversight as much as convenience. Roughly 2,100 people — about 20 percent of Parliament’s staff — already use AI daily, and much of that runs through public tools with no institutional control. Parliament’s leaders want that activity pulled onto a vetted platform, away from the consumer chatbots that draft legislation is being fed into today.

Concern inside the institution has been building. At a meeting earlier in July 2026, several committee chairs warned that a growing number of MEPs and their assistants were leaning on AI to draft amendments and other legislative work, Politico reported. The tells, according to those officials: a surge in amendments needing translation, recognizable drafting styles, and occasional citations to laws that do not exist. Those failure modes feed a wider worry about AI in political work, where research has found chatbots can skew in consistent directions when asked to weigh real laws.

Inside EPGenAI Hub

The platform splits access into two tiers. Users can pick models that run inside Parliament’s own infrastructure — Meta’s open-weight Llama, OpenAI’s open GPT-OSS, and Mistral Small — or reach for more capable services hosted externally, including OpenAI’s ChatGPT and Anthropic’s Claude Sonnet. The local option keeps sensitive material on the institution’s own systems; the external option trades that containment for more power. EPGenAI Hub has already been shown to MEPs and staff at workshops pitched around working “smarter and faster.”

That design answers a specific problem. In February 2026, Parliament disabled the built-in AI features on MEPs’ tablets after its IT team concluded it could not account for how much data those tools sent to external clouds. A sanctioned platform with an on-premises tier is the constructive version of that same instinct: give people a safe channel rather than only telling them what they cannot use.

Rules, not just tools, are the other half of the story. MEPs are under no obligation to disclose whether an amendment, a parliamentary question, or a speech was produced with AI. Staff face stricter terms — guidelines dating to April 2024, now being redrafted, require them to label AI-assisted work and bar them from entering sensitive information into public systems. Committee chairs have asked Secretary-General Alessandro Chiocchetti to draw up guidance for lawmakers; he suggested the matter may belong with the constitutional affairs committee, which would have to amend Parliament’s rules of procedure to impose anything binding.

For Brando Benifei, the Socialists and Democrats lawmaker who led Parliament’s work on the AI Act, the stakes are institutional credibility. “Parliament has not been immune to AI slop and misleading AI-generated content,” he told Politico, arguing that MEPs “must lead by example as the AI Act’s transparency rules become applicable.” Irish Renew member Barry Andrews, who chairs the development committee, was blunter about the line he draws: drafting questions, speeches and amendments are “core democratic functions which voters elected us to perform.”

A sovereignty problem it can’t escape

The rollout carries an awkward subtext. It opens Parliament further to American AI providers at the very moment Brussels is trying to wean the bloc off U.S. technology. The institution replaced Google with the French search engine Qwant as its default in June 2026 and framed the switch around digital sovereignty and data protection; the AI platform runs in the other direction, at least for its most capable models.

Parliament seems aware of the contradiction. Its press service said the model lineup “is undergoing revision, especially in light of availability of open-source and EU sovereign models from sovereign cloud providers recently contracted by the European Commission.” The American names on the menu today may not be permanent — a signal that procurement, not just capability, will shape what MEPs are allowed to use. It mirrors a broader European scramble for homegrown alternatives to Big Tech’s AI stack.

The episode is a small, concrete version of the bind the whole continent is in: Europe writes the rules for AI and, on paper, aims to set the global standard, but for the frontier capability its own lawmakers now want, it still reaches for models built elsewhere. Whether EPGenAI Hub becomes a genuinely sovereign tool or a managed gateway to U.S. systems will depend on what the Commission’s sovereign-cloud contracts deliver — and whether Parliament ever puts a disclosure rule behind the platform it is about to hand its members.