AI agents may revolutionize digital assistance, but their reliance on deep system access raises critical privacy and security concerns, according to Meredith Whittaker, President of Signal. Speaking at the SXSW conference last Friday, Whittaker underscored the risks posed by AI-driven automation.
During her keynote speech, Whittaker acknowledged the convenience AI agents bring to daily tasks, such as booking vacations, purchasing event tickets, updating calendars, and sending messages. However, she warned that enabling AI agents to perform these functions requires access to sensitive personal data, including web browsing history, GPS locations, credit card details, calendars, and messaging platforms.
“These systems need almost unrestricted access to our devices, essentially operating with root-level permissions,” Whittaker cautioned. “This means they interact with multiple databases, often without encryption, due to a lack of existing secure models.”
She further explained that this processing does not occur locally on users’ devices but instead takes place on cloud servers, raising significant concerns about data security.
“There is a deep-seated privacy and security issue at play,” Whittaker stated. “The hype surrounding AI agents is leading us toward a scenario where the boundaries between application layers and operating systems are blurred, potentially exposing sensitive user data.”
Given these concerns, Whittaker reaffirmed that privacy-focused messaging apps like Signal would not integrate with AI agents.
Her remarks follow previous criticisms of WhatsApp’s data collection practices. Last month, she accused the platform of harvesting excessive metadata, stating, “WhatsApp tracks who you communicate with, when, how often, and even your location. This data can be cross-linked with Facebook, Instagram, and payment information—Signal does not collect such data.”
In a separate interview, Whittaker highlighted the crucial role of private communication in various sectors, including journalism, human rights advocacy, military operations, and corporate confidentiality.
“When private communication is compromised, essential functions in society are at risk,” she said. “That is why platforms like Signal are vital in preserving secure and confidential interactions.”
As AI agents continue to evolve, industry leaders like Whittaker are urging tech developers and policymakers to address privacy risks before widespread adoption further erodes digital security.