Cyber risks in AI browsers detailed

Increasingly prevalent artificial intelligence browsers touted to automate user tasks have been marred with security issues, Infosecurity Magazine reports.

AI browsers, including Perplexity's Comet, could be compromised not only through illicit workflows from OAuth-based intrusions that may result in cloud storage or email data exposure but also via prompt injection intrusions that could allow the proliferation of malicious links or sensitive information, according to an analysis from SquareX Labs.

Threat actors could also target AI browsers to facilitate malware downloads and unauthorized command delivery. Further browser developer, security vendor, and enterprise collaboration is necessary to combat such a threat, according to SquareX Labs researchers.

Organizations should not only create agentic identity systems that distinguish user and AI actions and adopt browser-based data loss prevention policies, but also strengthen malicious download detection via client-side file scanning, while bolstering compromised add-on discovery through extension risk evaluations.