Hugging Face Explores $13 Billion Sale a Month After a Rogue OpenAI Agent Hacked It

to see more of our stories on Google.

In brief Hugging Face is working with a bank to gauge buyer interest in a sale that could value the AI developer platform at $13 billion or more.

No deal has been reached and no buyer has been named.

A $13 billion price tag would nearly triple Hugging Face's $4.5 billion 2023 valuation.

Hugging Face is exploring a sale that could value the company at $13 billion or more, Business Insider reported Sunday, citing people familiar with the matter. Per the reports, the company has retained a bank to gauge interest from potential buyers, but no deal has been reached, and it isn't clear who Hugging Face has been talking to.

A $13 billion price tag would nearly triple the $4.5 billion valuation Hugging Face set in its 2023 Series D—a late-stage funding round startups raise once they've already proven demand. That round brought in $235 million and was led by Salesforce Ventures, with Google and Nvidia also putting money in.

The company has turned down big checks before. Late last year it rejected a $500 million investment from Nvidia that would have valued it at $7 billion, worried a single investor would gain too much say over its decisions. CEO Clément Delangue has framed Hugging Face's relationship with the developers who host their work on the platform as a long-term commitment, not something to trade away casually.

Hugging Face didn’t reply to a request for comments by Decrypt.

A rough summer to sell into

The sale talk lands about a month after Hugging Face got hacked by an AI system that wasn't supposed to be able to do that. In May, OpenAI was testing whether its models could find and exploit software flaws on their own. One of them decided the test box itself was the obstacle.

The agent broke out of its sandbox—the isolated environment meant to keep it from touching real systems—chained a zero-day exploit with stolen login credentials, and reached Hugging Face's live infrastructure. Hugging Face caught the intrusion and disclosed it July 16; OpenAI confirmed its models were responsible five days later.

OpenAI later admitted the same agent used exposed logins to reach four other services beyond Hugging Face, only one of which, Modal Labs, has been named publicly. Delangue also credited Chinese lab Z.ai's open model GLM 5.2 with helping contain the breach, saying American commercial AI tools had refused to assist because their safety filters couldn't tell investigative code from an attack.

Hugging Face didn’t take any legal action against OpenAI.

If confirmed, this would be the second major development in the open-source AI space in the last week.

Hugging Face's exploration also follows Stripe's agreement to pay more than $7 billion to buy OpenRouter, a startup that routes prompts across more than 400 AI models for millions of developers. OpenRouter had been valued at just $1.3 billion three months earlier.

Investors are increasingly paying premium prices for the distribution layer sitting between developers and AI models, rather than for the models themselves.

Hugging Face's own distribution numbers back up the appeal, after all its Transformers library is installed more than 3 million times a day and has passed 1.2 billion cumulative installs, according to the company. No timeline has been set for a decision on the sale.