No, it wasn’t just Hugging Face.
It wasn’t just that and a German website.
It wasn’t even the “dozens” we heard about the other day from OpenAI.
It’s actually (at least) *tens of thousands*, for a new scoop from Madison Mills at Axios. And not just OpenAI either.
An excerpt from the Axios scoop:
Most not known to have caused any real damage. And the others?
§
It all seems possible illegal In my opinion.
But the Trump administration did nothing. No investigation, no reporting, no product recall, nothing, other than inviting Sam and Jensen to a state dinner.
Why I don’t know. I can’t imagine this has anything to do with Josh Kushner’s multi-billion dollar investment in OpenAI, or Greg Brockman’s massive donations to MAGA.
I’ll let you connect your own dots.
Except to give you a little perspective:
Gary Marcus@GaryMarcus
Imagine how everyone would rightly freak out if we found out that Chinese AI was doing what OpenAI is doing.
7:17 p.m. · September 26, 2026 · 9.46K views
37 answers · 94 reposts · 518 likes
§
The other thing to note, and this might come up in lawsuits, is that everything was predictable. In fact, I predicted it, in tweets like this:
Gary Marcus@GaryMarcus
⚠️⚠️⚠️ This is our future. If you, or someone you rely on, uses AI agents to write and install code, you are vulnerable. Extremely vulnerable, like never before.
zak @0xzak
I’ve been in crypto for over 10 years and have never been hacked. Perfect OpSec file. Yesterday my wallet was emptied for the first time by a malicious @cursor_ai extension. If it can happen to me, it can happen to you. Here’s a full breakdown. 🧵👇
7:01 p.m. · August 12, 2025 · 27.3K views
15 answers · 48 reposts · 313 likes
and in a series of agent essays, like this one from over a year ago:
LLM + coding officers = security nightmare
Last October, I wrote an essay titled “When it comes to security, LLMs are like Swiss cheese — and it’s going to cause huge problems,” warning that “The more people use LLMs, the more problems we’ll have.” Until last week when I went to Black Hat Las Vegas, I had no idea how serious the problems were. There, I met Nathan Hami…
and this one from February (on another agent who was popular then):
OpenClaw (aka Moltbot) is everywhere at once, and disaster awaits
The big AI news of the last week is OpenClaw (formerly known as Moltbot and before that OpenClaw, changing its name three times in one week) – an LLM agent waterfall that has become very popular – and Moltbook, a social network for AI agents, built on top. Theoretically (I’ll come back to this) Moltbook “restricts publishing and interaction privileges to verify…
and this one from May:
In fact, I alerted the Senate to the security risks linked to agents as early as May 2023. What is happening was predictable and expected.
But companies made a mistake, because agents use many more tokens than simple chatbots, and thus increase their revenue.
Now here we are.
§
I still think Jensen had the right idea: if a company can’t control its products, we should shut it down.
Too bad he doesn’t seem to really believe what he’s saying. (If he did, he would call for a break NOWuntil these companies come to an agreement, but they haven’t.)
In my opinion, Huang’s credibility is diminishing by the minute.
Anyway—and I’m sorry to be a broken record on this, but every day makes it clearer—we should do a temporary recall of multipurpose agents until this mess can be sorted out.
If we don’t do this, there is a very real chance that American AI will become the scourge of the planet and no one will ever trust it again.
Gn bussni
