Some agentic AI browsers come with major cybersecurity risks, UW study finds
A UW team studied seven popular agentic AI browsers and found that four create ways for malicious actors to bypass a fundamental cybersecurity protocol called the "same-origin poli…