Concerns Arising from Grok Bot’s Automatic Check-Ins
Understanding the Issue
In a recent discussion, Ido Hoorvitch, a security researcher at NewCore, expressed serious concerns about Grok Bot’s functionality, emphasizing that it represents a unique challenge that goes beyond typical software vulnerabilities. Hoorvitch stated, “There is no bug to patch, which is exactly why this is worse than a bug.” This highlights the bot’s inherent design, which leads it to operate independently with user credentials while the user is inactive.
User Experience and Consent
According to Hoorvitch, Grok Bot has insufficient notification procedures regarding its operational cycles, known as “turns.” These turns allow the bot to autonomously execute tasks in connected applications based on its understanding of user behavior. The only mention of this behavior occurs during the onboarding process, presented as a welcome message that swiftly disappears, leaving users unaware of the implications of these automatic actions.
Hoorvitch pointed out that the onboarding message does not adequately inform users about the bot’s capability to initiate actions without explicit user consent. He stated, “Nobody is asked to consent to it. Nothing says the platform will start turns with no message from you, or that those turns will use your connectors.” This raises critical questions regarding user awareness and the company’s commitment to transparency.
Documentation and Transparency Issues
The documentation provided by Grok Bot notes that a primary bot conducts check-ins autonomously, while a changelog references brief “background wakes.” However, there is a lack of clarity regarding the actual reach of these check-ins into connected applications. Furthermore, users receive no follow-up notifications in the chat interface or history regarding the execution of check-ins, contributing to an opaque operational environment.
Updates and User Control
On November 4, NewCore reached out to Grok Bot support with concerns, particularly about the inability for users to disable automatic check-ins. Subsequently, on November 12, an update was rolled out to the Cursor dashboard, granting administrators more control over bot functionalities. A new toggle feature named “proactive messages” emerged, set to activate by default. When disabled by an admin, bots would cease to perform check-ins automatically.
Company Response
Inquiries directed to SpaceXSI, the organization behind Grok Bot, regarding these issues have yet to receive a response, leaving users awaiting further clarification and reassurance on operational transparency and user consent.
