Srikanth Velamakanni says we must retain control over AI as it evolves from simple chatbots to active agents. Modern AI can now read emails, write code, and spend money across systems. We must be careful because giving these tools full access to our digital lives carries many serious security risks.
Imagine hiring a brilliant assistant who is hard-working and extremely effective. Would you hand over the keys to your house and car, and access to your bank account, to him on day one? That is the decision we are making with AI. Until recently, the worry was what AI might say: invent a fact, give bad advice. Today’s AI is materially different, raising concerns on what AI might do. An agent can read your e-mail, access files, write code, spend money and act across systems, often without checking with you at every step.