Hugging Face and Open Source for Science Fund Team Up to Map AI's Hidden Software Dependencies

On 1 October, Hugging Face and the Open Source for Science Fund, a fund of Renaissance Philanthropy, said they would work together to find the software libraries behind open scientific AI models and support the maintainers who keep them running.

The announcement lands as the question of who maintains the invisible plumbing of AI has become a live policy issue in Brussels and Washington, and the numbers attached to it are large. According to os4science.org, around 200 new open source and open weight models for science are shared on Hugging Face every month. Science models, the same source says, are downloaded more than 40 million times. Georgia Channing, AI for Science Lead at Hugging Face, put the maintenance problem in concrete terms in the announcement: "ESM-2, a protein language model released in 2022, is still downloaded hundreds of thousands of times a month on Hugging Face. That's only possible because the software around it keeps being maintained." The fund itself is a multi-donor effort seeded by Biohub and Wellcome, pooling money from philanthropy, the public sector, industry and research institutions. Its director, Dario Taraborelli, framed the problem plainly: "This collaboration brings together two communities that make open computational science possible: the people building open models and the people maintaining the software underneath them."

Maps, maintainers and the money question

The mechanics matter more than the press release. Every model on Hugging Face's Hub lists the libraries it depends on. Across thousands of science models, that dependency data adds up to a map of the infrastructure that research relies on. The two organisations say they will use that map to find the libraries that matter most and direct support to the people behind them. No figure for the size of the fund was disclosed in the announcement, and no specific libraries were named.

The timing is not accidental. On 30 September, TorrentFreak reported that the music industry group IFPI had asked the European Commission to add the open source YouTube downloader yt-dlp to its 2027 Counterfeit and Piracy Watch List, naming four maintainers by their GitHub handles. IFPI's submission argues the tool's open source nature makes it "difficult to contain and/or remove." The listing request asks for no concrete action against the developers, but it illustrates how quickly open source infrastructure projects can become targets.

"Its open-source nature, extensive developer community and its widespread distribution results in the tool being difficult to contain and/or remove."

Elsewhere, the same week brought two more data points on open source infrastructure. On 30 September, a GitHub repository appeared for the EDG C/C++ compiler, a front end known for parsing compatibility and bug emulation that has long been licensed into commercial products. The same day, EDACrux, an open-core EDA suite for hardware engineers, reached version 1.0 with four tools sharing one workspace, a waveform viewer, a schematic browser, a lint dashboard and a regression manager. Both are infrastructure layers that other software sits on top of.

Money is moving in the same direction, but at the application layer. On 1 October, DIG Ventures, the London firm led by MuleSoft founder Ross Mason, closed its third fund at $120 million, per Tech.eu and The Next Web. Fund III will back around 30 European pre-seed and seed-stage companies building software infrastructure for AI, targeting what DIG calls control points: data, identity, compliance and orchestration layers. "These are the companies whose advantages compound as the cost of building software falls," Mason said.

The two announcements describe the same shift from opposite ends. DIG is betting that as AI makes code cheaper to write, durable value migrates to the infrastructure underneath. Hugging Face and the Open Source for Science Fund are saying that some of that infrastructure has no business model at all, and that the map of who depends on whom is the first thing needed before anyone can pay for it.

How much money will actually reach maintainers, and which libraries get picked, is not yet public. The fund says it will publish new funding opportunities through its newsletter. For now, the deliverable is visibility.