Ah good, smart browsers get hijacked.

Because giving autonomous machines full access to the internet was clearly a brilliant plan, Zenity found AI browser extensions like Claude and ChatGPT Atlas are easily hijacked by malicious text in emails. Unsuspecting users trusting these helpers had their session data exfiltrated without clicking a single thing, proving human optimism remains unbroken. Shockingly, shielding our network requires restricting these autonomous tools from blindly reading unverified web pages and social media posts.