Spain's Sandbox IA — the EU's First National AI Regulatory Sandbox Under the AI Act
Spain
Under Royal Decree 817/2023, Spain became the first EU country to run a live AI Act regulatory sandbox: a 2025 …
Chile · Santiago · See the Chile profile · See the Santiago profile
Evidence: Observational / pre–post Top 15% 80/100 · Ask Evidence Copilot about this practice
Chile's SERNAC built a real-time alert system using NLP to scan 1,700-1,800 daily complaints for anomalies in company behaviour, cutting detection of systemic market abuses from weeks to under 24 hours — work that won a 2025 World Bank/ICPEN award.
SERNAC, Chile's National Consumer Service, built its Real-Time Consumer Protection Alert System (Sistema de Alertas en Tiempo Real) inside the Market Intelligence Unit (UIM) of its Department of Studies and Intelligence (DEI). The system ingests consumer complaints as they arrive and applies statistical anomaly detection to flag unusual patterns in company behaviour, triggering early coordination of protection action.
To detect systemic market abuses and trigger early coordination of consumer-protection action faster than manual review would allow.
According to the OECD's 2026 report on generative AI experimentation in government, SERNAC enhanced this pipeline with generative AI and natural-language processing at scale, parsing 1,700 to 1,800 daily citizen complaints, and produces a companion output called the Special Alert Bulletin for Claims (BEAR).
The OECD cites this as reducing the time needed to detect systemic market crises from several weeks down to under 24 hours. The system won first place in the 'Innovative Initiatives in Market Research' category of the 2025 Consumer Protection Contest, run jointly by the World Bank and the International Consumer Protection and Enforcement Network (ICPEN), presented 3 October 2025 in Santo Domingo. Chile's Ministry of Economy separately recognised SERNAC in April 2026 with its own PAEI award.
Evidence is corroborated across an OECD government report and multiple independent Chilean sources, but none discloses the underlying AI model, publishes false-positive/false-negative rates, or provides an independent technical audit of the anomaly-detection pipeline.
National / regional programmes
Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.
Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.
Where this practice's information was retrieved from, and when.
Spain
Under Royal Decree 817/2023, Spain became the first EU country to run a live AI Act regulatory sandbox: a 2025 …
United States of America
Under OMB Memo M-24-10, US federal agencies (except DOD and intelligence) must inventory their AI use cases yearly and publish …
Albania
Albania's labour inspectorate replaced its manual 'Matrix of Penalties' with MIRA, a nine-algorithm ML risk tool that now drives 70% …
Spain
In January 2025 the Basque Government published a public catalogue of algorithms and AI systems used by its administration, covering …
Open full copilot Grounded in cited practices — always check the sources.