BrokeIt - Daily AI News · All episodes: ↗

Hugging Face Hijacked, AI Hype Bill Due, Google Search Traffic Down 40%

2026-11-18 · 8 min

Listen · Apple Podcasts Listen · Spotify

Stories covered

Transcript

Intro

Ivy: …and with one expired domain name, tens of thousands of developers were suddenly running malicious code.

Marcus: This is Marcus.

Ivy: And this is Ivy.

Marcus: And you're listening to Broke It, your daily AI news for Wednesday, November 18th, 2026. We have a packed show for you.

Marcus: We're starting with how a popular AI model on Hugging Face got hijacked. It's a classic supply-chain attack with a very modern twist.

Ivy: We're also asking if the AI bubble is about to pop, as some of China's biggest tech CEOs warn the hype is way ahead of the value.

Marcus: And we've got new data showing Google's AI Overviews are absolutely tanking search traffic for publishers. The numbers are shocking.

Hugging Face Model Hijacked Through Expired Domain

Marcus: So, that Hugging Face hijack. This is a wild one. Researchers found a popular model, Kokoro AI, and noticed its creator's domain had expired.

Ivy: So they just… bought it? For what, ten bucks?

Marcus: Exactly. They bought the domain, reset the password for the associated email, and boom—they had full control of the Hugging Face account.

Ivy: And this wasn't some tiny, unused model. It had over 26,000 downloads. The researchers replaced it with a proof-of-concept that would steal SSH keys. This is… [sighs] this is so predictable.

Marcus: Predictable, but also a huge wake-up call. The whole open-source ecosystem is built on this web of dependencies. It's how we move so fast.

Ivy: It's built on a web of trust. And that trust is being systematically exploited. We treat models from platforms like Hugging Face like they're vetted, but they're just user-submitted code. It's the Wild West.

Marcus: But that’s the point! It’s on the community to find and fix these things. Which is what happened! These were white-hat researchers who disclosed it responsibly.

Ivy: This time. What about next time? How many other popular models are tied to a personal Gmail account and a GoDaddy domain that's about to expire? It's a systemic risk.

Marcus: Okay, so what's the fix? Lock everything down? Kill open-source collaboration?

Ivy: The fix is verification. Platforms need to enforce better security hygiene. Two-factor authentication that isn't tied to a recoverable email, warnings on models linked to personal domains… it’s not rocket science, Marcus.

Marcus: [beat] Right. So the takeaway for developers is clear: vet your dependencies. Check where the models you're pulling are actually coming from. You can't just `pip install` trust.

Ivy: And if you're a creator on one of these platforms, for God's sake, enable 2FA and auto-renew your domains.

The Bill for AI's Hype Is Coming Due

Ivy: Alright, let's talk about the money. Specifically, the giant pile of it being set on fire in the name of AI.

Marcus: [laughs] That's one way to put it. We've got a report from the South China Morning Post covering warnings from some heavy hitters—the CEOs of Baidu and Xiaomi.

Ivy: Right. Baidu's Robin Li is saying flat-out that there are too many foundation models being built, calling it a huge waste of resources. He thinks the real value is in applications, not in reinventing the wheel a hundred times.

Marcus: I mean, he's not wrong about needing applications. But isn't that how innovation works? A thousand flowers bloom, and then a few win out? You can't know which foundation model is best on day one.

Ivy: But the gap between hype and return on investment is getting absurd. Xiaomi's CEO said it could take years to see commercial success from their AI efforts. This is a classic bubble, Marcus: massive investment chasing future promises that may never materialize.

Marcus: Is it a bubble, or is it just the cost of entry for the next technological revolution? The internet boom looked like a bubble, and yeah, Pets.com went bust, but we also got Amazon and Google out of it.

Ivy: And in the meantime, thousands of people lost their jobs and their savings. The human cost of these hype cycles is real. The article mentions the 'huge gap' between what AI can do and what's profitable to do.

Marcus: So what's the alternative? We all just... stop? Stop building, stop investing, because it's too expensive and risky? We'd be ceding the future.

Ivy: No, you invest smarter. You focus on real-world problems instead of chasing AGI or building another chatbot that can write a poem about a pirate. The warning here is that the bill for all this hype is coming due.

Marcus: So the bottom line is, be wary of the hype. If you're at an AI startup, it might be time to ask some hard questions about the business model.

Ivy: And if you're an investor, maybe look past the buzzwords and toward companies that are actually solving a customer's problem. Profitability is not a dirty word.

Google's AI answers reducing search traffic by 40%

Marcus: And let's get to Google. A new report from RevScope finally puts a number on something we've all been feeling: the impact of AI answers on search.

Ivy: And it's a big number. Brace yourself.

Marcus: The study estimates that sites are seeing a forty percent drop in organic search traffic whenever a query triggers an AI Overview.

Ivy: Forty. Percent. Let that sink in. That's not a small dip, that's an extinction-level event for businesses built on SEO.

Marcus: Wow. I knew it was bad, but I didn't think it was that bad. So Google's AI is summarizing the web, and nobody's clicking through to the actual websites anymore.

Ivy: Why would you? The answer is right there. Google is consuming its own ecosystem. It's strip-mining the open web for data to train its models, then using those models to kill the traffic that keeps the open web alive.

Marcus: Okay, but isn't this just the next evolution of search? We went from ten blue links to knowledge panels and featured snippets. This is just the logical next step: the answer engine.

Ivy: [sarcastic] Oh, a 'great evolution'. What happens in five years when the creators and publishers that Google's AI relies on have all gone out of business because they have no traffic?

Marcus: The model will just... have all the information from the 'before times'.

Ivy: Information gets stale, Marcus! The web is a living thing! Or, it was. This is creating a future where new information is locked inside walled gardens or isn't created at all.

Marcus: So for anyone with a website, the reality is stark: the old rules are dead. SEO as we knew it is over.

Ivy: It means you can't depend on Google anymore. You need to build a direct relationship with your audience. Newsletters, communities, apps... anything that doesn't rely on a search engine as the middleman.

Marcus: Alright, let's recap the big stories.

Ivy: A popular Hugging Face model was hijacked with an expired domain, exposing a massive flaw in the AI supply chain.

Marcus: Some of China's top tech execs are warning that we're in an AI investment bubble, with hype far outpacing real-world returns.

Ivy: And Google's AI answers are cratering search traffic by as much as forty percent, threatening the whole online content ecosystem.

Marcus: And before we go... since it's November, I asked an LLM for a creative Thanksgiving recipe.

Ivy: Oh no. What did it give you? Marshmallow-brined turkey?

Marcus: [laughs] Close. It suggested a 'Cranberry and Wasabi Gravy'. It promised a 'zesty and surprising kick' to the meal. I think I'll stick with the family recipe.

Ivy: A surprising kick indeed. A kick right out of the dining room. Do not build that.

Marcus: That's our show for today! Thanks for listening to Broke It.

Ivy: We'll be back tomorrow with more. Until then, go check if your domains are on auto-renew.

This show is made with AI: the hosts’ voices are synthetic and the scripts are AI-assisted. Every story links to its original source.