Diritto intelligente – Issue N. 12

Summer 2025 marked a turning point for AI governance in Europe. With the European Commission’s approval of the GPAI Code of Practice, the AI Act has moved from principle to playbook.

Voluntary on paper, the Code is fast becoming a market benchmark: a way to demonstrate transparency, safety, and IP respect - while earning regulatory trust and reducing uncertainty. Early adopters will gain an advantage; late movers should expect tougher questions.

The EUIPO’s study on generative AI and copyright lands at the perfect moment. It underscores the fragility of today’s opt-out mechanisms for text and data mining and points toward interoperable technical standards, watermarking, and - ultimately - a licensing market that fairly compensates creators. Without that, friction between model developers and rights-holders will persist.

Meanwhile, the Commission’s GPAI guidelines clarify who is a provider, what counts as a GPAI model, and when systemic-risk obligations bite from 2 August 2025. Documentation, downstream disclosures, copyright policies, and training-data summaries won’t be nice-to-haves; they will be operational necessities with real enforcement behind them.

AI’s societal impact is equally present on the risk side. In insurance, AI supercharges both fraud detection and fraud itself - from ghost-broking portals to deepfaked identities. The AI Act’s transparency duties, combined with sector rules and coordinated supervision, can help - but only if institutions and industry invest in detection, evidence-grade audit trails, and public education.

Trust is also being rewritten in the fine print. Terms and Conditions and privacy notices are becoming reputational assets. Dense, opaque language now fuels backlash, especially where AI training is concerned. The path forward is plain-language drafting, layered summaries, and proactive communications that align legal accuracy with user comprehension.

Finally, the LegalTech market is growing fast, yet success hinges on process-first execution. Few organizations measure AI ROI rigorously; many still buy tools before mapping bottlenecks. The winners will be those who diagnose pain points, align technology to outcomes, and scale what measurably works.

The throughline of this issue is simple: AI is no longer “emerging” - it is restructuring law, markets, and accountability. Leadership now means building governance that is defensible, explainable, and human-centric.

Enjoy the issue.