Case study ·
Endesa uncovers electricity fraud in 16 Barcelona restaurants using AI and influencer promotions
Original title, in Spanish: “Endesa destapa fraude eléctrico en 16 restaurantes de Barcelona gracias a una IA y a la promoción de unos influencers”
- Publication
- Xataka (Spain)
- Author
- Isra Fdez
- Published on
- 9 October 2026 · read in the page structured data
Summary
In Barcelona, Endesa detected electricity fraud in sixteen restaurants using an AI that cross-referenced consumption data with Instagram videos posted by influencers paid to promote these establishments. Technicians at e-distribución, Endesa's network subsidiary, used a machine learning and deep learning algorithm to spot the inconsistency between the size of the premises, their apparent popularity on social media and their low billed consumption, revealing illegal connections to the public lighting supply. According to Endesa's press release of May 2026, the company detected 72,700 cases of fraud in 2025, 52% of which came from industries and large consumers. The CNMC estimates that this fraud adds around 150 million euros to the collective electricity bill. The article also raises legal questions about the use of social media data in these investigations.
Summary based on the full text of the article.
Read the article on Xataka →Figures cited
-
72,700 cases of fraud
electricity fraud cases detected by e-distribución (Endesa subsidiary) in 2025
Spain, e-distribución network, 2025
« su filial detectó 72.700 fraudes en 2025, la cifra más alta del último lustro y 1.700 más que el año anterior »
Source: Note de presse d'Endesa, mai 2026
-
320,000 cases
fraud cases closed between 2021 and 2025
Spain, e-distribución, 2021-2025
« Entre 2021 y 2025 se cerraron más de 320.000 expedientes y se recuperaron más de 3.750 GWh »
Source: Note de presse d'Endesa, mai 2026
-
3,750 GWh
energy recovered through fraud detection, 2021-2025
Spain, e-distribución, 2021-2025
« Entre 2021 y 2025 se cerraron más de 320.000 expedientes y se recuperaron más de 3.750 GWh »
Source: Note de presse d'Endesa, mai 2026
-
52 %
share of fraudulent energy attributable to industries, businesses, large consumers and cannabis plantations
Spain, e-distribución
« el 52% de la energía defraudada corresponde a industrias, negocios, grandes consumidores y plantaciones de marihuana, y solo el 5% a pequeños suministros de menos de 3 kW »
Source: Note de presse d'Endesa, mai 2026
-
5 %
share of fraudulent energy attributable to small connections under 3 kW
Spain, e-distribución
« el 52% de la energía defraudada corresponde a industrias, negocios, grandes consumidores y plantaciones de marihuana, y solo el 5% a pequeños suministros de menos de 3 kW »
Source: Note de presse d'Endesa, mai 2026
-
60 %
share of fraud cases involving direct connections without a contract
Spain, e-distribución
« Cerca del 60% de los fraudes son enganches directos sin contrato »
Source: Note de presse d'Endesa, mai 2026
-
5,487 cases
fraud cases detected in Barcelona in 2025
Barcelona, 2025
« la capital registró 5.487 casos en 2025, 891 más que en 2024 »
Source: TotBarcelona
-
24,347 cases
fraud cases detected in Catalonia in 2025
Catalonia, 2025
« Cataluña sumó 24.347, un 7 % más »
Source: TotBarcelona
-
50 %
proportion of inspections revealing energy theft thanks to machine learning and deep learning
Spain, e-distribución, compared with 2017
« e-distribución ya usaba machine learning y deep learning para detectar anomalías y que, gracias a ello, encontró robos de energía en una de cada dos inspecciones, el doble que en 2017 »
Source: eEXcellence
-
150 million EUR
estimated extra cost of electricity fraud on consumers' bills
Spain, CNMC estimate
« Según estimaciones de la CNMC que recoge El Español, estas prácticas encarecen la factura eléctrica en unos 150 millones de euros »
Source: CNMC
Classification
- Country
- Spain
- Company size
- Large companies
- Industries
- Energy, utilities, environment · Tourism, hospitality, catering
- Companies
- e-distribución (deploys AI), Endesa (deploys AI)
- Keywords
- fraud detection, artificial intelligence, machine learning, energy industry, social media, electricity consumption
Entry written by Marsanne Studio from the cited article, with the help of an AI model (claude-sonnet-5). The rights to the article and its illustrations belong to Xataka and its author. Entry translated from French by Claude Sonnet 5.5.