This Week in Startups · Monday, September 14, 2026
Jason Calacanis refutes the idea that AI developers are 'forced out of the loop' regarding AI's capabilities and potential risks. He points to companies like Claude implementing safeguards, preventing users from performing certain actions, and highlights that developers retain the ability to build in safety measures.
“So this flies in the face of what Claude's actually doing. Have you ever tried to get Claude to download a YouTube video? It's like, I can't do that for you, sir. Uh, okay, I'd like you to order me this ticket on United. I can find you the flight, get your permission. You have to give me browser permission. Then you log in, give me permission to use your password. And then I'll put into your cart, but I won't check out for you.”
“Literally, they're building products that have multiple safeguards in it so they don't get blamed for booking your ticket for you. Inadvertently.”
“So, yeah, for right now, what it does feel like is the AI is not capable of even doing everything you ask it to, let alone world domination. And we are capable of putting safeguards into it.”