Brave Browser is testing a new agentic AI feature, currently in its experimental phase. Emphasizing caution, Brave highlights the security risks associated with AI browsing, such as indirect prompt injections that could compromise user data. Unlike traditional models, Brave’s AI operates within an isolated profile, protecting users’ regular browsing data. The browser employs a secondary model to ensure that AI actions align with user intent, mirroring safety protocols used by Google Chrome. This alignment checker does not access website content, preventing potential prompt injections. Users retain control over AI tasks and can pause or stop the AI at any time. Brave maintains a strict no-logs, no-retention privacy policy, ensuring that user data is not used for training. Currently, agentic AI browsing is available through Brave’s Nightly channel, accessible via an opt-in setting. Explore the latest in AI browsing with Brave’s innovative features today!
Source link
Share
Read more