GPT@EC at the European Ombudsman — Guardrailed Generative AI for Case-Handling Support

The Office of the European Ombudsman, which investigates maladministration across EU institutions, has launched a pilot use of GPT@EC — a secure generative-AI tool built by the European Commission's Directorate-General for Digital Services on Microsoft Azure OpenAI infrastructure, offered to EU bodies for shared use in ancillary tasks.

Staff may use GPT@EC to research applicable law and case law, summarise and analyse large or complex documents and datasets, and support administrative drafting and communications. The Office's published rules state that 'AI is not used to take decisions, make recommendations, assess admissibility, determine outcomes, or prioritise complaints' — those functions remain exclusively with human staff, who must verify all AI output before use.

Governance safeguards are unusually detailed for a young pilot: a registered Data Protection Officer record, a completed data-protection impact assessment, a Service Level Agreement keeping Ombudsman data inaccessible to European Commission staff, a public AI-use notice, and a contractual guarantee that no processed data is used to train or improve the underlying models. As of the 2025 Annual Report, the Office (75 permanent staff, roughly 8 contract agents, 18 trainees) describes anticipated gains in speed and quality but publishes no measured figures on case-handling time, backlog, or output quality.

The European Commission's own digital-strategy office lists GPT@EC as part of a wider set of GenAI pilots trialled across EU public administrations, suggesting a path to reuse beyond this one office — but this remains a governance-first, evidence-later deployment.