The most powerful artificial intelligence companies in the world have made a sweeping public promise: Their technology will benefit all of humanity. Both Anthropic and OpenAI are now incorporated as public benefit corporations; according to Anthropic, this means that their directors are supposed to “balance the financial interests of the stockholders with…the responsible development and maintenance of advanced AI for the long-term benefit of humanity.” OpenAI has also committed to “making intelligence a tool that everyone can benefit from.” The third major competitor in the A.I. space, Google, remains a full for-profit corporation, but its A.I. division, DeepMind, has similarly published principles foregrounding social benefit and fairness. 

When A.I. companies commit themselves to “the long-term benefit of humanity,” they are setting a goal that Catholic social teaching has stressed for over a century: that the goods produced by human labor and ingenuity carry social obligations, and that those with power bear responsibility for ensuring that the benefit of those goods reaches everyone. As Pope Leo XIV’s encyclical “Magnifica Humanitas” reminds us, the principle of the universal destination of goods means that “the earth’s goods…are given by God to the entire human family to sustain the lives of all, and that every person has an inherent right to the use of such goods, both now and in the future” (No. 65).

Leo more specifically addresses artificial intelligence when he writes: “Data is the product of many contributors and should not be treated as something to be sold off or entrusted to a select few. It is necessary to think creatively in order to manage data as a common or shared good” (No. 108).

The first test of this challenge is simple, inexpensive and already overdue: A.I. companies should license full-featured versions of their tools to public libraries across the United States for free.

In 1731, Benjamin Franklin and his local discussion club formed the first modern library by pooling their books and making the collection available to all members. At that time, books were expensive, but their widespread availability was necessary, in Franklin’s estimation, for democracy to flourish. Two and a half centuries later, when the internet emerged as the defining information technology of the late 20th century, public libraries again served as the locus through which information and the expensive tools that harnessed it were made available to the public.

Collecting books in 1731 was a painstaking process. Adding internet access to libraries in the 1990s required a decade of infrastructure investment—new cables, new terminals, federal subsidy programs. A.I. requires none of this. The libraries already have the computers. The companies already have the software. All that is needed to bring A.I. tools to the public is a licensing agreement that allows the use of “pro-tier” A.I. models on library computers.

There are two reasons why this financial obligation should fall to A.I. companies rather than to library budgets or government grant programs. First, these companies built their tools on a commons they did not create and do not have exclusive right to: the accumulated writing of the open internet, digitized books, journalism, Wikipedia and the everyday text of hundreds of millions of people who gave no consent and received no compensation. When private actors extract enormous value from a shared inheritance, they incur an obligation to the community that produced it. Providing A.I. access to the public is not charity. It is fair compensation. 

Second, these companies have exacerbated an asymmetry that justice demands we rebalance. Through the capitalist marketplace, A.I. tools have been made available to those with the money to pay for them and denied to those most vulnerable in society. Health insurers use A.I. to generate claim denials, while patients appeal by hand; employers screen applicants with A.I. tools that are not disclosed to job seekers; debt collectors and consumer lenders use A.I. to assess, pursue and negotiate with borrowers, while the people on the other end of those transactions generally cannot.

Free tiers of all major A.I. platforms do exist, but they have low usage limits, run on older models, and omit the document analysis and extended reasoning that define what these tools can actually do. Pro-level capability costs between $20 and $200 per month—which can be prohibitive for an unemployed worker, a student navigating financial aid or a family working through an immigration process or debt dispute without a lawyer. 

“Magnifica Humanitas” also reminds us of the principle of solidarity. Leo writes: “Solidarity arises precisely when we decide not to remain indifferent to what happens to our neighbor but instead to transform unavoidable bonds—economic, cultural and technological—into paths of sharing, cooperation and mutual care” (No. 74). Solidarity is not sympathy; it is structural accountability. When a private actor creates a harmful asymmetry, the obligation to remedy it belongs to that actor. And when that actor has made commitments of public benefit part of their mission, this should be an easy lift. 

Public libraries are present in virtually every county in the country, with trained staff who make access meaningful rather than merely nominal—institutions of the common good in the most concrete sense. In March, the AI Center for Civic and Social Good opened at the Dr. Martin Luther King Jr. Library in San Jose, Calif., a partnership between the city, San Jose State University and Adobe that offers free A.I. tools and training to residents, students and small businesses, provided in part by OpenAI.

Now we need to scale this solution. The presidential Executive Order 14277 and the White House’s “Winning the Race: America’s AI Action Plan” identify the nation’s 123,000 library buildings as exactly the infrastructure this moment requires. But it is the public benefit A.I. companies that can and should fulfill the promise. 

Library licensing of full-feature A.I. models is the first test of whether Anthropic’s pledge of “A.I. for the long-term benefit of humanity” is a real statement of purpose or just a line in a press release. The libraries are ready. The need is clear. The A.I. companies have the means. All that remains is for them to decide whether they will fulfill their public commitments.

Matthew J. Gaudet is a technology ethicist in the San Francisco Bay Area and co-author of Encountering Artificial Intelligence: Ethical and Anthropological Investigations and Eight Theories of Justice: Perspectives From Philosophical and Theological Ethics.