Google is strengthening its security measures to protect users as it prepares to introduce automated features on Chrome, according to a recent report by TechCrunch. The automated capabilities being developed by various browsers aim to streamline tasks like ticket booking and shopping, but they also pose security risks such as potential data or financial loss.
In response to these challenges, Google has outlined its security strategy for Chrome, emphasizing the use of observer models and user consent for actions. In a preview back in September, Google revealed its forthcoming automated features and the accompanying security protocols.
To ensure responsible automated actions, Google has implemented a User Alignment Critic powered by Gemini. This critic scrutinizes planned tasks generated by the planner model and prompts a reassessment if they do not align with the user’s goals. Notably, the critic model only reviews metadata, not actual web content.
Moreover, Google is employing Agent Origin Sets to prevent unauthorized site access by agents. By restricting access to read-only and read-writeable origins, the browsing agent is limited to certain sections of a webpage, enhancing security and minimizing cross-origin data leaks.
Google’s vigilant approach extends to monitoring page navigation through URL checks using an observer model, ensuring a comprehensive security framework for Chrome’s automated features.
Source: TechCrunch
Leave a Reply