AI allows hackers to identify anonymous social media accounts, study finds | AI (artificial intelligence)
https://www.theguardian.com/technology/2026/mar/08/ai-hackers-social-media-accounts-study
Publish Date: 2026-03-08 10:00:00
Source Domain: www.theguardian.com
- AI technology, particularly large language models (LLMs), has made it easier for malicious hackers to identify anonymous social media accounts.
- Researchers demonstrated that LLMs could successfully match anonymous online users with their known identities by analyzing the information they posted.
- The study stresses the need for a fundamental reassessment of online privacy as LLMs enable cost-effective and complex privacy attacks.
- An example given in the study involved an AI correlating a user who mentioned “Dolores park” and a dog named Biscuit to a verified identity.
- The research flags potential misuse scenarios, such as governments surveilling dissidents and hackers conducting highly personalised scams.
- AI surveillance is a developing field that many experts find alarming due to its potential for sophisticated online information synthesis.
- Experts warn that the lowered barrier for more developed attacks makes malicious activities easier, requiring only publicly available language models and an internet connection.
- There is significant concern about commercial use of AI for de-anonymizing, which could cause individuals to be wrongly accused of actions.
- LLMs can sometimes fail to accurately link accounts, leading to misidentification and misuse of shared data from various sources beyond social media.
- While LLMs show potential for de-anonymizing records, they are not perfect and may not succeed in situations with insufficient information or many potential matches.
- To address these issues, scientists recommend institutional changes and increased individual precaution in sharing online information.
