Incident 836: La province du Sichuan est en proie à de nombreux rapports fabriqués par l'IA sur des catastrophes et des crises
Description: Au Sichuan, des outils d'IA auraient été utilisés pour créer et diffuser de fausses informations dans plusieurs villes, notamment des histoires inventées de toutes pièces de glissements de terrain, de tremblements de terre, d'affrontements, d'accidents et d'incidents sanitaires. Ces rumeurs, alimentées par l'IA, ont induit le public en erreur, semé l'inquiétude et la confusion, et mis à rude épreuve les autorités locales. Les auteurs présumés cherchaient à s'engager sur les réseaux sociaux et à générer des profits grâce à des contenus appâts à clics. Les autorités du Sichuan ont finalement infligé des sanctions administratives aux auteurs.
Editor Notes: Incident 836 presents an editorial challenge in that it is a composite of 10 discrete AI incidents that point to the possibility of other harm events (see Incidents 834 and 835 as well). "Moumou" and "Mou" refer to partially anonymized names in Chinese media. The following is a reconstruction of each incident outlined in this collective incident report; the majority (but not all) incidents were in Sichuan Province: (1) 11/20/2023 – Fabricated Chengdu Bus Accident: Chen Mou reportedly altered news of a bus crash in Chongqing to claim a similar incident occurred in Chengdu, alleging fatalities and severe impact. (2) 11/28/2023 – False Chengdu Construction Accident Report: Zhu Mou created a fake story about a construction accident in Chengdu’s Longquanyi District, falsely reporting multiple worker fatalities due to an electrical incident. (3) 12/27/2023 – Underground Pipeline Rupture Hoax in Xi’an: Wang Mou and Liu Mou allegedly spread a rumor about a broken underground heating pipeline in Xi’an, using AI to combine information with misleading images to attract attention. (4) 01/12/2024 – Fabricated Abandoned Infant Story in Guiyang: Zhou Mou reportedly used AI to create a story claiming a four-month-old infant was abandoned in Guiyang’s Huaxi District, falsely presenting a distressing social issue. (5) 01/23/2024 – Fabricated Landslide Fatalities in Yunnan: Yang Moumou allegedly spread a fake story claiming that a landslide in Yunnan resulted in eight deaths, using AI-generated information to amplify the incident’s impact (see Incident 835). (6) 04-05/2024 – False Banner and Ceremony Reports in Chengdu: Wang Mou allegedly created fake news about a banner encouraging second births, along with a fictional ceremony in Chengdu’s Pidu District, using AI to create convincing but false visuals. (7) 05/11/2024 – Heart Attack Incident in Xichang: Ji Moumou allegedly created a story about a man who reportedly died of a heart attack following a sex-related incident, spreading the rumor online to gain views. (8) 05/18/2024 – Bazhong Cash Transport Standoff: Shang Moumou allegedly fabricated a report about a standoff involving a cash transport vehicle in Bazhong, claiming a police-citizen confrontation for dramatic effect. (9) 06/09/2024 – Exaggerated Xide County Earthquake Report: Luo Moumou used AI tools to fabricate a story about an earthquake in Xide County, amplifying the damage and claiming casualties, despite the actual quake occurring in Muli County with minimal damage (see Incident 834). (10) 06/24/2024 – Elderly Abandonment Story in Chengdu: Zhang Mou reportedly created a false story of a 90-year-old man abandoned by his children, claiming he was found begging in Chengdu’s Xinjin District, using AI-generated visuals.
Entités
Voir toutes les entitésAlleged: Unknown deepfake technology developers et Unknown AI developers developed an AI system deployed by Zhu Mou , Zhou Moumou , Zhang Moumou , Yang Moumou , Wang Mou , Shang Moumou , Luo Moumou , Liu Mou , Ji Moumou et Chen Mou, which harmed Sichuan general public et Sichuan authorities.
Statistiques d'incidents
Risk Subdomain
A further 23 subdomains create an accessible and understandable classification of hazards and harms associated with AI
3.1. False or misleading information
Risk Domain
The Domain Taxonomy of AI Risks classifies risks into seven AI risk domains: (1) Discrimination & toxicity, (2) Privacy & security, (3) Misinformation, (4) Malicious actors & misuse, (5) Human-computer interaction, (6) Socioeconomic & environmental harms, and (7) AI system safety, failures & limitations.
- Misinformation
Entity
Which, if any, entity is presented as the main cause of the risk
Human
Timing
The stage in the AI lifecycle at which the risk is presented as occurring
Post-deployment
Intent
Whether the risk is presented as occurring as an expected or unexpected outcome from pursuing a goal
Intentional
Rapports d'incidents
Chronologie du rapport
Gou Chun, journaliste de couverture
Ces dernières années, les fonctions de génération de texte et d'images de l'IA ont été utilisées par certains criminels pour modifier de fausses informations, abaissant ainsi le seuil permettant aux crimi…
Variantes
Une "Variante" est un incident de l'IA similaire à un cas connu—il a les mêmes causes, les mêmes dommages et le même système intelligent. Plutôt que de l'énumérer séparément, nous l'incluons sous le premier incident signalé. Contrairement aux autres incidents, les variantes n'ont pas besoin d'avoir été signalées en dehors de la base de données des incidents. En savoir plus sur le document de recherche.
Vous avez vu quelque chose de similaire ?
Incidents similaires
Did our AI mess up? Flag the unrelated incidents
Incidents similaires
Did our AI mess up? Flag the unrelated incidents