AI amplifies the risk of social media overexposure: inferring data that was never shared
Everyday exposure on social media is no longer observed only by other people: it is also seen by artificial intelligence systems capable of linking large volumes of data. Engineer Julio Domínguez says what has changed is not the data but the ability to extract meaning from it, with multimodal models that process text, images, audio and video. A study by ETH Zurich using Reddit profiles showed that large language models could infer location, income or sex, with up to 85% accuracy on their first guess and 95.8% with three answers.
In 40 seconds
- AI enables automated analysis of data that previously required teams of analysts and predefined goals
- Multimodal models cross-reference text, images, audio and video, and can infer information never provided
- An ETH Zurich study using Reddit profiles achieved up to 85% accuracy in deducing personal attributes
Read the full story at EL PAÍS
IASummary generated by artificial intelligence from a story by EL PAÍS. It may contain errors: the original source prevails.
Comments
Be respectful. Comments reported by several people are hidden until a moderator reviews them.
You do not need an account to read, only to comment. Sign in with a code sent to your email or with your Google account. We only show the name you choose, never your email.