This immediately removes AI assistants (e.g., Copilot/ChatGPT/Claude) as an approved workflow for drafting, summarizing, or searching parliamentary correspondence on managed devices, pushing staff back to non-AI tooling or offline processes. Vendors seeking to sell AI productivity tools into EU institutions now face a higher bar for on-prem/EU-hosted processing, auditable data-handling, and guarantees that prompts and documents are not retained for model training. The move also signals that cross-border legal access risk (including U.S. lawful demands) is being treated as an operational security constraint for institutional deployments, not just a compliance issue.