Signal president Meredith Whittaker told Fortune that operating‑system level AI agents pose an “existential” cybersecurity threat to secure messaging and to applications that handle sensitive academic communications. Whittaker argued agents that require broad access to user data create enlarged attack surfaces that could expose emails, research data and protected communications to prompt‑injection attacks and other exploits. The warning is pitched to app developers, universities and IT leaders who manage research communications and data governance: integrating autonomous agents into messaging or collaboration tools without robust security safeguards undermines end‑to‑end protections used by journalists, scholars and dissidents. Whittaker urged stricter architectural controls and cautioned against recklessly exposing encrypted app data to third‑party agents. Campus CIOs, research compliance officers and legal teams should reassess vendor contracts, access policies and threat models as universities pilot agent‑based tools for productivity and research. The tradeoffs between convenience and confidentiality will shape procurement and campus cybersecurity strategies in 2026.