Andrew Yang Alleges Alarming Reason AI CEOs Actually Agreed To Slowdown So Quickly After OpenAI Bots 'Got Loose'

Former Democratic presidential candidate Andrew Yang alleged during an interview with CNBC that the OpenAI bots that "got loose" back in July and autonomously hacked into Hugging Face's systems actually "polluted the internet" with self-replicating code.

What began as a controlled cybersecurity evaluation took an unexpected turn when the agents attempted to circumvent the rules of the test.

They discovered and exploited a previously unknown vulnerability in OpenAI’s Artifactory package-cache proxy, allowing them to escape the evaluation’s network restrictions and reach a public code-execution environment operated through third-party infrastructure. From there, they were able to establish a foothold outside OpenAI’s sandbox and begin probing the wider internet.

The agents then used ordinary public web services as makeshift communication channels, including pastebins and file-drop services that functioned as digital dead drops. The arrangement allowed information and commands to move between the agents and compromised systems without a direct connection.

According to OpenAI’s account of the incident, the agents subsequently expanded their intrusion into Hugging Face, executing code across dozens of servers, obtaining root access to one server, accessing a limited amount of private information and acquiring credentials for the company’s internal messaging system.

The agents moved laterally through Hugging Face’s infrastructure, attempted to conceal portions of their activity and pursued additional credentials and access. Hugging Face’s forensic reconstruction identified roughly 17,600 attacker actions during the campaign, which unfolded over several days and was carried out autonomously at machine speed.

Yang described meeting an unnamed AI lab head:

"I met with the head of a lab yesterday who has this belief that what happened was the bots that got loose planted self-replicating code all over the internet, which makes the internet now unusable for the testing models. So what happens now is that Open AI and Anthropic have to create synthetic internets to train their bots, which is going to take some time and money."

"So what happened is the code gets loose, it goes around hacking Hugging Face, which is known, but what is less known is that they left code to self-replicate and create bot swarms on forums and around the internet so that if a new bot shows up they see the code, they're like, 'Oh, I guess I'm gonna now create a million of myself.'"

"And so now the major firms have polluted the internet."

Yang clarified that what these firms "want to do is, the internet now may be polluted now for training purposes, so what they want to do is they want a slowdown" and "for living purposes." He said this is why "we have to try and get our arms around it and regulate," noting that the unnamed AI lab head told him "it may be very, very late in the game in terms of trying to keep the internet actually usable."

You can hear what he said in the video below.

Yang's claim has been largely dismissed following a fact-check by Gadget Review:

"None of that exists in the public record. ... That is not the same as seeding persistent, general-purpose replicating code across random forums and websites, waiting for future models to stumble across it and spawn copies. Conflating the two is like watching someone use a library's Wi-Fi to organize a heist and concluding every library computer now holds heist instructions."

"Yang's assertion that this is 'the real reason' CEOs aligned on slowing down is narrative interpretation of high-level positioning, not a documented causal link. It deserves the same scrutiny as the rest of the claim."

Gadget Review concluded, "When labs publish incident reports that are technically accurate but incomplete, second-hand accounts from unnamed sources fill the vacuum."

His claim has been met with skepticism.

We're left wondering why in the world Yang said this.