I gave Meta’s Muse and OpenAI’s Dots my passwords for a week
Meta’s and OpenAI’s new agents can streamline your life. Artificial intelligence researchers helped me test how far to trust them
Two AI agents, named Mark and Sam, were used to handle various tasks for one week by a journalist. Mark, created by Meta, and Sam, by OpenAI, are designed to streamline daily life activities such as making purchases, sending emails, and organizing errands. Mark had access to the journalist's email, bank accounts, social media, and Spotify, while Sam had limited access and was blocked from some platforms.
Both agents appeared as conversational text message chains with friendly avatars. The journalist found Mark more successful in completing tasks, such as finding and purchasing a couch, while Sam struggled with access to certain accounts. The agents' personalization capabilities were also tested, with Sam suggesting a used book the journalist had previously requested and Mark recommending a concert ticket and a lamp.
However, concerns about privacy, user control, and potential misuse of these AI agents' access to personal information were raised.
Written by urgent.news from Scientific American's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
