MIT Professor: The Most Dangerous AI Application on the Internet Just Went Viral…

What happens when AI agents stop waiting for instructions and start acting on their own? Julian sits down with Ramesh Raskar and Maria Gorskikh for an urgent conversation about the rise of proactive AI and the emerging “agentic web”. This isn’t about chatbots answering questions anymore, it’s about autonomous agents that can send emails, execute tasks, coordinate schedules, and interact with other agents without constant prompting. As open-source tools like OpenClaw make it easier for developers to spin up agents, we may be witnessing a shift from centralized AI models to a world where anyone can run their own personal AI. But with that power comes serious risk, from malicious “skills” and prompt injections to identity spoofing, financial manipulation, and large-scale bot collusion. Without the right infrastructure, guardrails, and accountability, this new frontier could quickly become chaotic. And yet, the upside is enormous. They explore a future where billions of people, from fa