#44: Wozniak YouTube scam, Unprompted AI agents BTS, agentic shopping is here

This week we’re talking about trust: how it’s earned, how it’s abused, and how it’s being rewritten by machines.

From a verified YouTube scam targeting Steve Wozniak, to AI agents making real-time security decisions, to shopping bots quietly reshaping eCommerce flows, automation doesn’t just change the game. It changes who’s playing.

Let’s get into it.

NATE'S TAKE - AUGUST 12, 2025

Top Three This Week

1. Steve Wozniak Wants Platforms to Do More to Stop Scams
2. AI Agents Don’t Wait for Prompts Anymore - And That Cuts Both Ways
3. Pinterest Says Agentic Shopping Is Years Away. We’ve Already Used It.

1. Steve Wozniak Wants Platforms to Do More to Stop Scams

Apple co-founder Steve Wozniak is still fighting a YouTube scam that used real footage of him to promote a fake Bitcoin giveaway. The video was hosted on a verified channel and claimed Woz would double your crypto. One woman lost nearly $60,000.

He and his wife flagged the fraud repeatedly. YouTube left it up. Five years later, their lawsuit is still stuck behind Section 230 protections, which protects platforms from liability for user-generated content.

This is impersonation fraud with a trust halo. Real video. Fake frame. Verified badge. No friction. Platforms aren’t just hosting this stuff, they’re legitimizing it.

If your platform shows verified badges, allows user videos, or runs influencer content, ask yourself: what are the fraud implications when those signals are hijacked? And how fast can your team respond when trust is being abused?

2. AI Agents Don’t Wait for Prompts Anymore - And That Cuts Both Ways

Artistgndphotography | E+ | Getty Images

Cybersecurity teams are starting to deploy AI agents that don’t just detect and alert, they act. Tools like Google’s new async coding agent Jules are part of a growing wave of “autonomous teammates” that can now monitor logs, reason through threats, quarantine inboxes, and even restrict compromised accounts without waiting for a human to ask.

The shift isn’t hype. It’s pressure. Phishing emails aren’t sloppy anymore. Attackers are moving faster, and defenders are offloading the grunt work to machines that can keep up.

A Gartner poll says 1 in 4 CIOs already have AI agents running, mostly behind the scenes. But the next wave is coming for fraud teams too. These flows don’t look risky. There's no friction, no flags, and no time to second-guess.

3. Pinterest Says Agentic Shopping Is Years Away. We’ve Already Used It.

Pinterest CEO Bill Ready says agentic shopping - where AI agents shop on your behalf - is “still a long way out.” He believes most people aren’t ready to give up that much control.

But we tried it. I prompted an agent to buy shoes for my co-founder Patrick. It hit Nike first (blocked), bounced to other sites until it got through to one, signed up for a discount code, applied it, and completed the purchase.

Shopping is changing. Fast. And underestimating this shift is dangerous. Agents are users now. If they’re able to make purchases on your platform, they’ll blend right in.

For fraud teams, this means clean sessions aren’t always safe sessions. For CX and trust teams, it means the checkout flow you’ve optimized for people is already being navigated by bots. And for marketplaces and social platforms: if you’re building discovery-first shopping experiences, these agents are already looking for exploits.

Agentic commerce is a new attack surface and a new customer segment. Treat it like both.

===

That’s all for this week! For more insights, follow us on LinkedIn or X, and if you want to learn more about what we do, visit www.specprotected.com.

‍