Signal president says agent-based AI has 'profound' security and privacy issues

At the SXSW conference, Signal CEO Meredith Whittaker expressed serious reservations about the use of agent-based artificial intelligence, TechCrunch reports.

She emphasized that such systems can pose a significant threat to user privacy and security, as agent-based AI requires access to a large amount of personal data to perform its tasks.

Whittaker explained that to perform tasks like booking tickets, managing a calendar, or sending messages, an AI agent needs to have access to a user’s browser, credit card information, messaging apps, and other databases. She emphasized that the processing of this data is likely to occur in the open on cloud servers, which poses serious security risks.

The Signal CEO also noted that integrating such agents into messaging apps like Signal could undermine user privacy. She warned that the current AI paradigm, which is based on massive data collection, has potentially dangerous consequences. In her opinion, the pursuit of a "magic bot" that solves all life's needs could further weaken the protection of users' data.