Who Are You, Really? Identity in the Age of AI
Adam Clark Estes recently wrote something very good over at Vox (syndicated via Yahoo Tech) about AI-powered scams and how to protect your family from them. He’s a longtime tech reporter who’s spent years warning his parents about online scams. Then an AI fake nearly fooled him, too. If it can happen to him, it can happen to any of us. His story, and his practical advice, are well worth your time.
Here’s what stuck with me. Lots of us use AI tools in our day-to-day lives now, and so do the bad guys. They clone a grandchild’s voice to call grandma asking for money. They test endless variations of a phishing email until one convinces the accounting clerk to hand over the company’s banking credentials. Which means the oldest question in email, “Who is this, really?”, matters more than ever.
Email has a good answer to that question: authentication. DMARC lets a receiver check DNS to verify that a message claiming to come from a domain was actually authorized by that domain’s owner.
So what happens when AI agents start acting on our behalf? Valimail co-founder Alexander Garcia-Tobar tackles that in a recent op-ed for CIO. His argument: before asking what an AI agent can do, ask “Whose agent is this?” He proposes borrowing directly from DMARC, using DNS to verify an agent’s identity before deciding what it’s allowed to do. And once you know who an agent belongs to, you can start building reputation and accountability, just like we do with email senders today.
I agree. Whether it’s a phone call from “grandson,” an email from “the bank,” or an AI agent claiming to represent your vendor, the first question is the same: Who are you, really? Email figured out how to answer that. Now it’s time to answer it for everything else.
