Building on open source AI? Hugging Face’s policy chief on the costs, risks, and trade-offs

Open source AI models are moving into the mainstream as enterprises, governments and startups chase the promise of lower costs, tighter control, and more flexibility. In a new episode of TechCrunch's Equity podcast, host Rebecca Bellan sat down with Irene Solaiman, Chief Policy Officer and SVP of Strategy at Hugging Face, to examine what startups need to weigh before building on open models — from tool and infrastructure choices to the trade-offs shaping open AI's future.

The conversation starts with the economics: why organizations of every size are switching to open AI models to cut costs, and, conversely, when it makes less sense to reach for the most powerful models available. Solaiman also addresses a structural puzzle of the open ecosystem — who actually makes money when the models themselves are free to use, and whether the companies producing them can capture the value they generate.

A second thread runs through U.S.-China competition. Solaiman argues the U.S. may need deep-pocketed backers such as Nvidia to stand up an open frontier model capable of competing with China's, and spells out the stakes if Chinese models end up as the global default for builders everywhere.

Security closes out the agenda. Recent attacks on open AI platforms have sharpened a difficult question for the community: how do you keep an open ecosystem secure without effectively closing it off? For Hugging Face, which hosts one of the largest public repositories of models and datasets, the tension between openness and defensive posture is not abstract — it determines what developers can download, fine-tune, and ship.

Solaiman brings both a policy and a strategy lens to these questions in her dual role at Hugging Face, positioning the discussion for startups weighing vendor lock-in against total cost of control. What to watch next hinges on several open questions the episode surfaces: whether capital at the scale of Nvidia's actually materializes behind an American open frontier model, how openly licensed Chinese models reshape default model selection outside the U.S., and whether security measures adopted after the recent attacks can hold the line without fragmenting the open ecosystem. Each answer will bear directly on how startups pick models, where they run them, and what they end up paying for — in dollars or in autonomy.