Gartner advises organizations to block the use of AI browsers. According to the agency, these new-generation browsers pose risks that are difficult for most organizations to manage.
The advice was reported on by The Register. In particular, the combination of AI functions and autonomously acting agents poses a structural security problem.
The warning comes from a recent Gartner advisory. In it, analysts note that AI browsers are designed with ease of use as a priority. Security is of secondary importance. This means that these browsers deviate from what organizations try to enforce through their security policies. Browsers that visit websites, fill out forms, and perform actions within logged-in sessions significantly increase the attack surface.
Gartner classifies AI browsers as solutions that integrate an AI side function. This function can summarize, search, or translate web pages, combined with agent functionality that can perform tasks independently. In practice, this means that large amounts of user data, such as open tabs, active sessions, and browsing history, are sent to cloud-based AI services. According to Gartner, without strict central configuration, this quickly leads to unwanted exposure of sensitive information.
AI browsers undermine compliance
In theory, organizations can limit the risk by thoroughly assessing the underlying AI services and determining whether their security meets internal standards. But even if that assessment is positive, Gartner says caution is still warranted. Employees must be aware that content they view in the browser may end up on external AI platforms without their knowledge. This makes the use of AI browsers problematic in scenarios where confidential or regulated data is processed.
When organizations conclude that the AI backend used is not sufficiently trustworthy, Gartner recommends not allowing AI browsers within the IT environment. The risk lies not only in data leaks, but also in the behavior of autonomous agents. These are susceptible to manipulation via indirect prompt injections, can draw incorrect conclusions, or be misled into navigating to phishing pages, potentially resulting in the loss of login credentials.
Although some technical measures can limit the risk, such as blocking email functionality or preventing storage, Gartner considers this insufficient to make AI browsers safe to use. According to the analysts, their use requires extensive risk analyses, strict policies, and continuous monitoring. Even then, the list of permitted applications is expected to remain limited, while the management burden increases.