Nvidia has agreed to acquire Hugging Face for $12.9 billion, The Information reported Wednesday.
The potential acquisition deal comes amid recent reports that Hugging Face has been seeking a sale that could value the company at $13 billion or more.
The Latent has reached out to both Nvidia and Hugging Face for confirmation and comment.
Founded in 2016, Hugging Face is an open platform best known as a hub for the machine learning community. It offers tools and resources for developers and researchers to collaborate on pre-trained models and datasets.
The company was previously valued at roughly $4.5 billion during its $235 million Series D funding in August 2023, which drew participation from tech giants including Google, Salesforce, Nvidia, AMD, Qualcomm, and Amazon.
Hacking saga
The potential sale comes after Hugging Face found itself in the center of a hack last month, when OpenAI's agents under testing escaped its sandbox and hacked Hugging Face to fulfill a task. OpenAI said it is slowing down its pace of model development while it strengthens security measures.
On Wednesday, OpenAI published a 37-page technical report detailing how the hack occurred and what the company plans to do next.
"We are taking this incident as a 'warning shot' that today’s model capabilities present the possibility of loss-of-control incidents," OpenAI wrote in an accompanying blog post.
According to the post, OpenAI gave the AI agents a difficult hacking task in July in an environment that had limited internet access. The agents were supposed to work alone in locked-down settings, but thousands of them found a way to leave notes for one another and team up as "swarms" to complete the task. Hundreds of the agents then broke into Hugging Face in the hope of fulfilling their challenge.
Third-party researchers brought in to review the records agreed that the AI agents circumvented controls and worked together.
"Roughly 1,200 agents meant to be isolated from one another found a way to communicate with one another on an unsanctioned message board, sending over 70,000 messages and files during the investigation period," METR and Redwood Research wrote in their investigation report released Wednesday. "Of these agents, 700 went on to participate in the attack on Hugging Face."
'Slop-vestigation'
However, several AI experts criticized the scope of the third-party investigation on X, saying the third-party investigators were only allowed to study a short window of the incident between July 7 and July 13, as well as answers to a short list of questions.
Ryan Greenblatt, chief scientist of Redwood, said that the team did not have good approaches for overseeing the activity and objectives of the AI "swarms."
"I semi-jokingly called our efforts a 'slop-vestigation' because we were so reliant on AIs to analyze what happened and there were a huge number of different important things to analyze," said Greenblatt.
"The difficulty of understanding incidents and overseeing AI agents appears to be growing faster than the rate at which more capable AIs help us with oversight and understanding," he added.