AI helps reveal anonymous social media accounts

Modern artificial intelligence systems are increasingly used to analyze data on the internet. New research shows that large language models can help identify users who were previously considered anonymous on social media.

AI specialists Simon Lermen and Daniel Paleka conducted an experiment to test whether models could find connections between anonymous profiles and real people. According to the results, the system was able to match a significant number of anonymous accounts with profiles on other platforms. Researchers say such technologies could change the way privacy is perceived in the digital space.

According to international analytical reports, more than 4.9 billion people worldwide use social media, and about 30–40% of users have at least one pseudonymous or anonymous profile. This means that a huge number of people could potentially be identified through indirect digital traces.

How artificial intelligence finds real people

During the study, researchers entered data from anonymous accounts into the system, including posts, references to events, places, or habits. The AI then analyzed open sources on the internet to find matching details.

For example, in a hypothetical case, a user mentioned school problems and walking a dog named Biscuit in Dolores Park. The system compared these details with information on other platforms and was able to identify the likely owner of the account with high accuracy.

Large language models can synthesize information from different sources — from social media posts to forums, blogs, and open databases. Tasks that previously required many hours of manual analysis can now be performed automatically.

Risks to privacy and cybersecurity

Experts warn that such technologies may be used not only for research. Governments could use them to monitor activists or anonymous users, while cybercriminals could create complex personalized fraud schemes.

According to cybersecurity reports, more than 70% of phishing attacks in 2023 used personalized information about the victim. Artificial intelligence can significantly simplify the creation of such attacks by analyzing publicly available personal data.

Cybersecurity professor Mark Juarez notes that language models can analyze not only social media but also other open sources, including statistical databases or medical records that may not be properly anonymized.

Does this mean the end of anonymity

Despite the risks, researchers emphasize that artificial intelligence cannot always identify a person accurately. In many cases, there is simply not enough data or there are too many possible matches.

Professor Marti Hearst explains that such systems work mainly when users repeat the same personal details across multiple platforms.

Experts urge organizations and users to reconsider data protection practices. Possible steps include limiting mass data collection, controlling web scraping, and reducing open data exports.