Users Love What Instinct's AI Assistant Can Do — But Its Sweeping Access Is Raising Flags
The short answer
A new AI assistant called Instinct is generating strong early reviews for its capabilities, but privacy and security observers are flagging serious concerns about the breadth of system access it requires and what it can do autonomously on users' behalf. Before adopting any AI assistant that requests access to your email, calendar, files, or business accounts, read its terms of service — specifically what data it can access, how it is stored or used, and whether your inputs can be used for model training.
- AI assistants
- privacy
- security
- data access
- terms of service
- small business
What happened
TechCrunch reports that Instinct, a new AI assistant, has drawn enthusiasm from early testers for its performance. That enthusiasm is paired with growing concern: the assistant is described as having sweeping access to users' systems, broad terms of service, and the ability to take actions on users' behalf — combinations that some early testers say come with uncomfortable trade-offs.
The report does not disclose the full specifics of Instinct's terms, but privacy and security are the central concerns being raised. The pattern echoes earlier debates around other high-capability AI assistants: the more an AI can do autonomously, the more access to your accounts and data it typically requires to function.
What it means for your business
Before adopting any AI assistant that requests access to your email, calendar, files, or business accounts, read its terms of service — specifically what data it can access, how it is stored or used, and whether your inputs can be used for model training. For small businesses that handle customer information, broad-access AI tools can create compliance exposure, particularly under state privacy laws or if you operate in healthcare, legal, or financial services. A practical rule: grant any AI tool the minimum access needed for the specific job you are hiring it to do, and revoke access to systems it does not need. Positive early reviews and strong capability alone do not eliminate security risk.
Source: TechCrunch