“Parliament has not been immune to AI abuses”: the EU is about to deploy its own AI platform, including tools from OpenAI, Meta, Anthropic and Mistral

European Parliament presents secure AI platform for legislative use
Officially sanctioned platform will allow MPs and staff to use AI models
AI errors and hallucinations have become a problem for the European Parliament

The European Union is considering introducing an officially sanctioned AI platform to offer European lawmakers a safer alternative to publicly available AI tools.

The platform, called “EPGenAI Hub”, will include the Meta, OpenAI, Anthropic and Mistral models designed for legislative purposes. Several Members of the European Parliament (MEPs) have expressed concerns about the growing use of AI within the European Parliament, which has led to the development of a platform that does not transmit sensitive information to public models.

Around 20% of Parliament’s current staff use AI tools in their daily work – around 2,100 people – which has contributed to some legislative documents containing AI-related errors and hallucinations.

EU to adopt new AI platform

“Parliament has not been immune to the abuses of AI and misleading content generated by AI,” said Brando Benifei, a socialist and democratic deputy. “MEPs must lead by example as the AI law’s transparency rules become applicable.” Benifei led the European Parliament’s work on the development of the AI law.

Currently, there are no regulations requiring MPs to disclose whether AI tools have been used to draft or edit legislative documents, but Renew’s Irish lawmaker Barry Andrews said there is clear evidence of AI being used within the European Parliament. “It’s their choice and I completely understand it given our workload,” he said. “But for me personally, these are essential democratic functions that voters elected us to do.”

However, there are guidelines for parliamentary employees, Policy stating that any material modified or produced using AI tools must include an “AI-assisted” disclaimer, and that employees are not allowed to introduce sensitive information into AI tools.

MEPs and European Parliament staff benefited from demonstrations of the new EPGenAI hub. The platform runs on Parliament’s own infrastructure and external hosting, enabling the secure use of a range of AI models such as Meta’s Llama, OpenAI GPT-OSS, Mistral Small, OpenAI ChatGPT and Anthropic’s Claude Sonnet.

Google logo on black background next to text