As neural networks continue to advance, they pose a significant risk: the potential creation of AI-generated digital personas that can persist even after a person's account or query history has been deleted. According to expert Yaroslav Seliverstov from the University 2035, this is made possible by the formation of a non-transferable digital footprint, where elements of a person's behavior, speech patterns, visual image, and social patterns can be retained in trained model parameters. Currently, AI systems are trained on vast amounts of data, including social media posts, which can lead to a future where deleting an account is no longer sufficient to erase one's digital presence. The ownership of these digital personas is also unclear, as they may be formed from distributed data from multiple sources. In response, Seliverstov predicts that governments will need to shift from regulating data storage to regulating AI memory, potentially introducing the right for individuals to request the removal of their digital patterns from training datasets and generative models, as well as requiring labeling and consent for the creation of digital duplicates. This raises important questions about system design and the technical architecture of AI systems, highlighting the need for data engineers to consider the long-term implications of their work and the potential consequences of creating persistent digital personas.

Source: RIA Novosti