A group of scientists has raised concerns about the safety of artificial intelligence-powered web browsers, warning that they are not yet secure enough for widespread public use. These browsers, often called 'agentic browsers,' are designed to perform complex tasks automatically, such as booking flights or filling out forms. However, researchers say that the very features meant to make them helpful also create serious security vulnerabilities.
The scientists highlight that these advanced capabilities can expose personal information. For example, the browsers might share sensitive data with third parties without the user's knowledge or consent. The risks stem from how the browsers handle tasks like logging into accounts or processing payments. According to the experts, the systems are still too immature to guarantee user safety.
The report does not specify the exact flaws but emphasizes that the technology has significant red flags. The scientists urge developers to address these issues before promoting such browsers to the general public. They recommend that users stick to traditional browsers for now.
AI-powered browsers are part of a growing trend where software acts on behalf of users. While convenient, the new study indicates that security has been overlooked in the race to add features. The researchers call for more rigorous testing and transparency to protect users' data.