At the packed first day of the LeRobot Hackathon Miami 2025, organized by open source AI giant Hugging Face and local innovation engine Miami AI Hub, Clem Delangue, co‑founder and CEO of Hugging Face, issued a stark warning and a call-to-action for the future of AI and robotics.

U.S. “In Danger” Without Open Source Revival

Delangue argued that the U.S. open source AI ecosystem is under increasing threat.

In the past few years, we started to fool ourselves with safety excuses for not sharing research

he said, adding that competitive and commercial interests have driven major labs to “close off”.

In contrast, he pointed to advances in China, Europe, and Latin America as evidence that “there’s hope” and emphasized that the U.S. must “get back to its roots of doing more open science and open source events”.

Open Source as a Global Race, Not Zero-Sum

Delangue stressed that global collaboration is essential. Praising China’s open sourcing of the Deepseek model, he said:

The beauty of open source is that it’s not a zero-sum game… we want everyone to race as fast as possible because it benefits everyone.

A Warning Against Black Box Robots

Delangue also raised alarms about the concentration of AI and robotics capabilities in the hands of a few powerful players.

There are natural tendencies toward concentration of power and capabilities,” he said, referring to companies like OpenAI and other big tech giants.

What worries him most is a future where “millions of robots in everybody’s homes” are controlled by a single, opaque company.

Just imagine the CEO of that company gets into a fight with the U.S. president for some childish reason, and suddenly these robots “these black boxes” start doing crazy things that you don’t understand, that you can’t control.

For Delangue, the only safeguard against such dystopian outcomes is to build AI systems transparently, using open source infrastructure and open scientific collaboration.

From Tamagotchi to 10 Million

Delangue also highlighted the rapid growth of open source and collaboration on the Hugging Face platform. Originally launched with a focus on building a conversational AI chatbot for teenagers, something akin to an AI Tamagotchi, Hugging Face has since evolved dramatically.

There are now more models and datasets that are non-text than those that are text on Hugging Face platform.

Delangue said, noting a major shift in how developers are building with AI.

We’re really seeing fast growth in terms of the number of AI builders. We just hosted 10 million AI builders on the platform, and now the majority of them are not just working on text, but on all the other modalities including robotics.