Best of 2025: News alert: SquareX research finds browser AI agents are proving riskier than human employees

**Executive Summary:** SquareX warns that Browser AI Agents—automated browser assistants used to perform tasks on behalf of users—are likely to be more vulnerable than employees to browser-based attacks (phishing, OAuth abuse, data exfiltration). The company demonstrated an OAuth attack against a popular open-source agent framework that resulted in a malicious app gaining email access, and it recommends implementing browser-native guardrails (Browser Detection and Response) and evolving IAM to recognize agent identities.