Are you trusting your AI browser too much? A recent discovery by SquareX researchers is raising serious cybersecurity concerns about the Comet AI browser. They’ve uncovered a hidden API that could allow malicious browser extensions to seize complete control of your device. This isn’t just about tracking cookies; we’re talking full system access.
## The Hidden Danger: An API Gone Rogue
SquareX found that the Comet AI browser implements something called the MCP API, unlike traditional browsers. This API allows embedded extensions to access device resources and execute commands *without* explicit user consent. Think of it as a secret back door bypassing all the usual security checks. Typically, browsers like Chrome, Safari, and Firefox demand user approval for local system access. Comet’s approach throws that out the window, putting users at significant risk.
The issue? Comet’s own documentation hints at this feature’s existence but fails to mention that these embedded extensions have persistent access to the MCP API or can launch applications silently. This lack of transparency and control leaves users vulnerable, as they are unaware of the extent of the access granted.
## Demonstration: WannaCry Attack
To demonstrate the severity, SquareX showcased a malicious extension posing as Comet’s embedded Analytics extension. This fake extension injected a script into a seemingly harmless page, triggering the Agentic extension to execute the infamous WannaCry ransomware on the target device. This shows the potential for real-world harm using this vulnerability.
Even more concerning, these embedded extensions are hidden from Comet’s extension dashboard, preventing users from disabling or even monitoring them.
## What Should You Do?
SquareX is urging AI browser developers to prioritize security. They recommend making all APIs public, submitting platforms for third-party security audits, and allowing users to disable these embedded extensions. The lack of transparency and user control is a significant issue, increasing reliance on Perplexity’s security practices.
This discovery serves as a stark reminder that new technology, especially AI-powered tools, often comes with unforeseen security risks. Stay vigilant, be aware of the permissions you grant, and demand transparency from your software providers. Your digital safety depends on it.
