Meredith Whittaker, president of privacy‑focused messenger Signal, warned that operating‑system level AI agents pose an "existential threat" to secure apps by expanding the attack surface for sensitive data. Whittaker told Fortune that agents require broad access to contacts, messages and credentials to operate—access that could be exploited via prompt‑injection and other attacks. The warning highlights cybersafety challenges universities face as campuses pilot AI agents for advising, research and administration, and raises questions about platform choices and data‑governance policies for protected research and student information.