The Tour d’ Horizon of Data Law Implications of Digital Twins
would vary. For instance, health and defense data are categorized as sensitive in comparison to data about a machine’ s health( such as in a Manufacturing DT). Further, it may not be guaranteed that the cloud service provider and the Data Processor may ensure security of the data hosted on cloud servers.
4.5.2 IMPACT
A data breach or data leak of the cloud may happen, which could impact the privacy of an individual( under Healthcare) or loss of proprietary information( under other Use Cases).
Cloud storage solutions may be deployed by the DT provider directly or may be procured from a third-party through a contractual arrangement. In both the cases, the liability for any breach / leak would be on the DT provider, being the Data Controller. Therefore, it becomes relevant for Data Controllers to contractually ensure that cloud service providers and Data Processors( who have access to the data stored on cloud) maintain adequate security, adopt safeguards and monitor mechanisms and report data breaches.
4.6 DATA FIDELITY
As mentioned above, DTs are data intensive and the accuracy of data used for the creation of these DTs is of critical importance. Inaccuracy in the Input Data or historical data may lead to inaccurate or erroneous outcomes.
4.6.1 POTENTIAL RISKS
In the Healthcare Use Case, a DT may be used for diagnosis and treatment. If such a DT is fed with inaccurate, incomplete or inconsistent records of patient data, it may result in faulty and incorrect diagnosis and treatment of the disease or predictions related to medicines to be administered. Inaccurate outcomes resulting from inaccurate, incomplete or inconsistent Input Data or historical data may cause harm to human health( in case of Healthcare Use Case) or may impede operational efficiency in other Use Cases.
AI forms an intrinsic part of the DT model and may be influenced by biases from the past. In the context of Healthcare, data may be subject to such bias which may impact the outcome. For instance, a study reported by MIT revealed that AI takes a“ short-cut” by basing the diagnosis on learnt demographic categorizations. 50 Furthermore, research indicates that discrimination in healthcare can persist through historical records of diagnoses made by individual doctors decades ago.
These biases, embedded in the data from 20 years ago, can inadvertently influence modern AI systems. Even if the discriminatory factors are no longer prevalent in society today, their
50
Trafton, A. Study reveals why AI models that analyze medical images can be biased, MIT News https:// news. mit. edu / 2024 / study-reveals-why-ai-analyzed-medical-images-can-be-biased-0628.
88 May 2025