One week ago, Hugging Face was celebrating the 3 million models hosted on its platform. © Hugging Face
Hacked in July by ChatGPTs, courted in August by buyers in the $13 billion range. For Hugging Face, the summer of all dangers is turning into a jackpot, and even its defenses have a story to tell.
The story could have ended with a spectacular hack, already recounted in our columns. It has now taken a far more lucrative turn: according to revelations published this weekend, Hugging Face reportedly hired a bank to evaluate takeover bids of around 13 billion dollars.The platform, founded in 2016 by three French entrepreneurs and now the central repository for open AI models, has not confirmed any deal, and no buyerâs name has surfaced.
The yearâs most profitable hack
The numbers tell a story no one saw coming. In the two weeks following the intrusion, the volume of data uploaded to the platform surged by 58%, according to a chart published by Clément Delangue, its CEO (the figure therefore comes from the company itself, to be clear). Meta followed suit by publishing on Hugging Face this month its first general-purpose open model since 2023, a return to the fold that serves as an endorsement. The French Ministry of Culture is also among the organizations that upload their resources there, a sign that the platform has long since moved beyond the circle of research laboratories.@@TAG@@
On the valuation front, the excitement heated up in three stages. The 2023 Series D valued the company at $4.5 billion; earlier this year, it turned down a $500 million Nvidia investment at a $7 billion valuation; now it is being courted at nearly triple that. There are benchmarks: Microsoft paid $7.5 billion for GitHub in 2018, and Stripe has just acquired OpenRouter for $7 billion. Delangue himself tempers the excitement by recalling his âlong-term responsibilityâ to a community that entrusts its models to him, leaving all doors open, including the option of saying no.
The detail that puts closed-source models in an awkward position
An episode of the counterattack, revealed this week by the American daily The New York Times, sheds light on the other side of the affair. To analyze the intrusion, Hugging Faceâs engineers reportedly first turned to Anthropicâs model; its guardrails reportedly interpreted the request as assistance for an attack rather than as a defense, bringing the maneuver to an abrupt halt. The team switched to an open-source model from Chinese company Z.ai, the famous GLM already encountered in this case, which enabled them to identify how to lock the intruders out.
Hugging Face, for its part, handles the anecdote with a delicate touch that reeks of courtesy. In its own account of the incident, the platform names no one and takes care to emphasize that the episode is in no way an indictment of the guardrails of hosted models. The lesson it draws is entirely practical: have a vetted model capable of running on its own infrastructure ready before the crisis. Still, the story could not be more timely for the current open-source crusade, and it partly explains why hosting open models has become an asset everyone wants to own. Meanwhile, OpenAI is holding back its own models for security reasons.
đđ» Follow tech news in real time: add 01net to your sources on Google, and subscribe to our WhatsApp channel.
Source: New York Times
- Google Play Store
- 4K Video Downloader Plus
- VLC Media Player
- Adobe Reader
- uTorrent
- WinRAR
- Xender
- Google Chrome
- CCleaner
- Mozilla Firefox
- Avast Free
- ChatGPT
- Microsoft 365
- Opera One
- Avast One
Download
- Google Play Store
- 4K Video Downloader Plus
- VLC Media Player
- Adobe Reader
- uTorrent
- WinRAR
- Xender
- Google Chrome
- CCleaner
- Mozilla Firefox
- Avast Free
- ChatGPT
- Microsoft 365
- Opera One
- Avast One
