Digital Identity Challenges in AI: Findings of Duplication in Google Knowledge Graph

The introduction of digital identity within the artificial intelligence (AI) ecosystem is now facing new challenges. As the consolidation of data on the internet becomes increasingly complex, the phenomenon of identity duplication in global databases has begun to spark scientific discussion regarding the effectiveness of algorithms in accurately mapping human entities. This issue has come to the fore following the discovery of a suspected anomaly in the Google Knowledge Graph, a massive database system that connects billions of pieces of information in the search engine and underpins various AI-based services. The finding was revealed by Muhammad Ari Pratomo, an Indonesian legal practitioner, after conducting an independent investigation into the validity of his digital footprint. In his analysis, Ari found two different data entry entities, or Knowledge Graph Machine IDs (KGMID), which are strongly suspected to refer to the same individual. This case serves as a real-world illustration of the entity resolution problem in computer science, the process by which automated systems determine whether data from various sources refers to a single entity. The existence of two KGMIDs for one person indicates that modern AI systems still have technical gaps in synchronising widely dispersed data. As a result, a person’s digital information can become fragmented and incomplete. Although it cannot be categorised as a permanent system error without official confirmation from the developer, this phenomenon demonstrates the limits of automation technology. The complexity of the internet makes the unification of identity from billions of data points a complicated matter, even for the world’s largest search engine platform. ‘If both truly represent the same individual, this condition can serve as an example of the challenges in the entity resolution process, namely how an AI system determines that diverse information from various sources indeed refers to a single person,’ said Muhammad Ari Pratomo, explaining the technical impact of the phenomenon. Among technology practitioners, the debate has now shifted to the extent to which external evaluation can influence algorithm refinement. Some parties assess that this duplication may be part of an unfinished data synchronisation process, while others see it as a signal of the need for fundamental improvements to digital identity recognition systems. This case confirms that no matter how sophisticated an automation system is built, the role of human oversight in testing the technology’s reliability remains crucial to creating a public database that is objective, valid, and free from information errors.