A new Swiss study has found that AI tools can successfully unmask anonymous social media accounts with up to 68% accuracy, threatening the foundation of online privacy.
Artificial intelligence tools can now unmask anonymous social media accounts at scale, matching pseudonymous users to their real identities with alarming accuracy, according to a new study from ETH Zurich. Researchers found large language models correctly identified up to 68 percent of matching accounts with 90 percent precision, “substantially outperforming” non-LLM methods like human investigation.
“Our results show that the practical obscurity protecting pseudonymous users online no longer holds,” the study warned. Lead author Daniel Paleka told The Independent the findings make it “very clear” that “if you keep posting under a pseudonym, keep quoting information about yourself,” AI tools will unmask you cheaply and quickly.
The researchers outlined grave implications: governments could surveil dissidents, corporations could hyper-target advertising, and attackers could launch personalized scams. Older and vulnerable users who leak more personal information face greatest risk. “Keep in mind that everything you post stays on the internet and can become the target of future models,” Paleka cautioned.