Would you trust an unsupervised 9 year old child with your work email, intimate text messages, and credit card? Everyone was warned long ago that AI agents are highly vulnerable to phishing emails with hidden malicious instructions. Giving an agent the ability to click links means it will click links, including dangerous ones. AI cant stop this either, so don't listen to any AI company or cybersecurity vendor claiming otherwise. AI can't detect a malicious email that looks legitimate. It will never detect a malicious URL unless its already classified as dangerous. this means it will never detect new dangerous URLs. Don't allow an AI agent to do anything you wouldn't trust an unsupervised 9 year old to do with your phone or computer. In 2015, my team built the world's first security chatbots that did the opposite. Ours were installed inside business tools and messaging services like Slack and Messenger. On Slack, our bot detected phishing scams, removed dangerous messages, banned the threat actor, and sent a DM to the account owner or admin to inform them what happened. It was my team's security bot that eradicated phishing on Slack for the entire crypto world in in 2017. Millions of people were interacting with chatbots way before they realize. Conversational chatbots aren't new. Ours didn't have AI, so they weren't intelligent, but allowing a chatbot to act on your behalf is obviously risky. They're not as intelligent as you might think and they're prone to making more mistakes than a human. Any security company claiming this is a new discovery is living under a rock or playing a PR game.