Meet Instinct: The personal AI assistant that works for us, at the cost of privacy concerns
Instinct, a new personal AI assistant from Silicon Valley, can handle all tasks for us, but is raising major concerns regarding privacy and security.
In the era of rapid AI development, a startup named Spear Street Technology, Inc.—led by Noah Shinn, a former researcher at Sierra and co-author of the AI agent research paper "Reflexion"—has developed a personal AI agent named Instinct, which is currently in private, invite-only testing.
Designed for proactive tasks via SMS or WhatsApp, the tool can manage and reply to emails, schedule calendar appointments, and even conduct transactions and online shopping on behalf of the user. It is backed by major investors including Kleiner Perkins and Conviction.
However, early testers and security analysts have begun raising major concerns. The first issue is "deep data access," as the system requires permissions to access Gmail accounts, messages, locations, audio, and on-screen behavior in order to genuinely perform tasks for the user.
Additionally, the Terms of Service include a "perpetual and irrevocable license" condition covering the use of personal data to train AI models. There are also security risks regarding prompt injection vulnerabilities, given the agent's high-level system control privileges, with some testers reporting instances where the system executed actions without explicit confirmation.
It reflects the double-edged sword of modern AI agents that can do more work on behalf of humans, but at the cost of exposing personal data and granting deeper device access—a major security challenge that all users must watch out for.