Studies Show AI Agents and Browsers Are a Hacker’s Perfect Playground

Red team tests show that enterprise AI agents and browsers are alarmingly easy to hack — from leaking CTO credentials to auto-filling phishing sites with sensitive data. Studies reveal over 3,000 exposed Copilot Studio agents and repeated failures of AI browsers like Comet. The verdict is clear: without security-first design, agentic AI will remain a hacker’s perfect playground.