Companies backing the foundation include Google and Salesforce as “core partners,” and Hugging Face, Instacart, AI startup Sakana AI and others as “supporters,” suggesting they have committed less money to the project.

Martin Tisné, the CEO of the AI Collaborative and France’s special envoy for public interest AI at the AI Action Summit, told Fortune he envisions many of these projects being around the creation of new public datasets that can be used to build AI models that serve the public interest.

He says the bottleneck for the creation of public goods in AI is not access to computing power, but rather data. Much of the data people would want to create models, to do things like help cure cancer or help people use AI to better analyze government policies, are not found on the internet. In many cases, governments have these data sets, but they are not easily accessible by the public. In other cases, governments are in a position to create high-value datasets that the private sector cannot create, although Tisne noted that these would have to be created and made available in “a secure and privacy-preserving way.”

Tisné said the new foundation would also look to support the development of smaller AI models that are targeted to specific use cases. Smaller models have multiple benefits in that they are easier to deploy safely, with developers more easily able to ensure that they do not amplify biases or engage in algorithmic discrimination. Smaller models also tend to use less energy to train and run. AI’s energy footprint and impact on water usage are becoming a key concerns as it becomes increasingly apparent that the widespread adoption of very large AI models could contribute to the consumption of more fossil fuels, exacerbating global warming, and worsen water shortages.

He said the foundation would fund efforts to create a “resilient open source AI ecosystem,” seeing openness as a way to ensure broad public access to AI technology. “The market is at a juncture where it could go either way,” Tisné said, explaining that AI could either concentrate power in a few firms with proprietary models or democratize the technology through open models.

Finally, Current AI will make grants around AI accountability, providing support for efforts to increase AI transparency, auditing of AI models and datasets, and public engagement over AI.

In addition to France, which is the primary government behind the initiative, it is being supported initially by eight other countries, including Germany, Finland and Switzerland, but also Chile, Kenya, and Nigeria, among others. “If you are not the U.S. and not China, you are more in the same boat [with every other country] than you think,” Tisné said.

Hugging Face CEO Clement Delangue and Instacart CEO Fidji Simo were among a group of ten prominent technology entrepreneurs and venture capitalists who released an open letter over the weekend calling for the creation of AI “public goods,” with the letter’s release timed to come just ahead of the foundation’s formal announcement on Tuesday. 

The main philanthropic funders, besides the Omidyar Network’s AI Collaborative, are the John D. and Catherine T. MacArthur Foundation, the Patrick J. McGovern Foundation, and the Ford Foundation. 

The initiative reflects France’s push to frame AI policy around opportunities rather than just potential risks. French President Emmanuel Macron released a statement saying that “by giving access to data, infrastructures and computing power to a large number of partners, Current AI will contribute to developing our own AI ecosystems in France and Europe.”