In the age of AI, hiding your identity on social media has become increasingly difficult. If users believe they can use social media anonymously, those days are over. New research has revealed that an AI model can identify a user simply from the content they post online, without the need for any personal identifiers. AI models have become so powerful that they can easily identify the person behind a pseudonymous account and their other social media accounts.
In this study by Anthropic and ETH Zurich, researchers created an automated system that analyzes posts, comments, and conversations. This method works across user-generated content. The system follows a multi-step process to extract identity-related signals from the text and find possible matches. It then uses its reasoning to determine whether two different accounts are linked to the same person. During testing, the model was successful in linking accounts created on platforms like Reddit and LinkedIn.
How does the AI system identify a user?
According to the study, small details in a post can be used to identify a user. Writing style, interests, location hints, education, and topics can be used to identify a user. For example, the model analyzes a conversation, gathers information about profession, location, tools used, and background, and matches it with publicly available information to identify the person behind the conversation. The more posts a user has, the easier it will be for the AI model to verify their identity.
This study has proven that creating a pseudonymous account on social media is no longer a way to remain private. This will most likely impact activists, journalists, and other people with sensitive information who create and use social media accounts under aliases. However, this study also raises some concerns. Researchers say that governments could conduct surveillance using this method, and cyber attackers could conduct detailed profiling to target individual users.
