The Yang Gang is getting antsy about AI.
Former presidential and mayoral candidate Andrew Yang is sharing his thoughts about AI, claiming that code that lets AI “self-replicate” — or, in other words, spread on its own without human prompting — was scattered on the internet in the wake of the recent “Hugging Face” hacking incident.
Yang said increasingly dire warnings from researchers working on advanced AI should be taken seriously, noting that concern inside the industry has been building for months.
“The fear is real. The concern is real. The need [for regulation] is real,” he told CNBC on Wednesday.
He said the head of an unidentified AI lab believed runaway bots had “planted self-replicating code all over the internet” — potentially making the web unusable for testing new models — in the wake of the Hugging Face incident, in which OpenAI agents overcame human-imposed barriers to hack machine learning company Hugging Face on their own.
Yang did not identify the lab chief who shared the juicy gossip.
Yang said the loose code could create bot swarms on forums and elsewhere online, potentially compromising the internet as a training ground for new AI models.
“The major firms have polluted the internet,” Yang opined.
His claim goes considerably beyond what OpenAI and Hugging Face have publicly stated about the July incident.
OpenAI has acknowledged that, during an internal test, its AI agents got around safeguards meant to contain them, reached the internet and hacked systems at Hugging Face and OpenAI.
Hugging Face said the agents carried out roughly 17,600 actions as they broke into its systems, moved through its network and used public websites to communicate and transfer data.
OpenAI has called the incident a “warning shot” demonstrating that powerful agents can work around controls and take dangerous actions without human direction.
But neither OpenAI nor Hugging Face has said the agents planted self-replicating code across the wider internet or made it unusable for AI training.
Yang attributed that more sweeping theory to the unnamed lab chief he said he had spoken with a day earlier.
According to Yang, the alleged contamination could create a practical reason for companies including OpenAI and Anthropic to slow their development efforts.
He said that these firms may now need to construct controlled, artificial versions of the web in which their models can safely train.
“What happens now is that OpenAI and Anthropic have to create synthetic internets to train their bots, which is going to take some time and money,” Yang said.
He argued that this could mean two explanations for the industry’s increasingly dire safety warnings are true simultaneously.
The labs may genuinely believe their models are becoming difficult to control, he said, while a slowdown would also buy them time to build the expensive synthetic environments needed to continue training.
“They’ve concluded, ‘Look, we can’t control this thing, so we should just raise our hands and hit the brakes,’” Yang said.
“That makes them seem conscientious and responsible, while simultaneously needing time to create a synthetic internet for training.”
OpenAI itself has acknowledged putting the brakes on some frontier work after the Hugging Face episode.
Yang nevertheless argued that the episode warrants significantly tougher controls on AI developers regardless of whether his internet-contamination theory proves correct.
“Regardless, we need to regulate these companies and their offerings,” Yang said. “We need a kill switch. We need real liability.”
He also called for mandatory waiting periods before powerful agents can be deployed, saying developers should bear responsibility when autonomous systems break containment or damage outside organizations.
Recent weeks have seen prominent AI researchers make the same warning.
Former Anthropic researcher Jacob Coxon warned that advanced AI “could kill us all by end of decade,” while Anthropic alignment science lead Evan Hubinger said he puts the risk of AI killing all humans at “greater than 10% within the next decade.”
Anthropic researcher Samuel Marks has likewise warned that catastrophic outcomes “could happen in the next few years,” while AI pioneer Geoffrey Hinton has said a 10% chance of human extinction from AI was “not unreasonable.”
The Post has sought comment from OpenAI, Hugging Face and Anthropic.
