Does Desktop AI Come With a Side of Risk?

The article warns that the arrival of desktop AI assistants (Microsoft 365 Copilot, Apple Intelligence, and Google's Project Jarvis) amplifies existing security risks—oversharing of information, insufficient access controls, and new attack surfaces such as prompt-injection attacks that can coerce agents into leaking data or performing fraudulent actions—and urges organizations to gain visibility, implement granular/time-limited permissions, and monitor agent use to mitigate these risks.