Workers and users need journalism that examines how AI affects jobs, public decisions and daily life—not only what companies say their systems can do. The Tarbell Center for AI Journalism aims to distribute at least $10 million in institutional grants by the end of 2027. That support can strengthen scrutiny, but only if readers can see how the money is allocated and trust that funders do not steer the reporting.
The $10 million is a target, not a sum already distributed. The test is not simply whether the nonprofit funds more AI coverage. It is whether its decisions are legible to the public and whether the journalists receiving support remain in charge of their work.
Grants can give AI reporting room to grow
Tarbell’s institutional program expands on earlier grants for individual stories. Semafor reported that those grants ranged from $1,000 to $20,000, with $400,000 in commitments over four years. The new program offers grants generally in the $50,000 to $500,000-plus range and says it may also make multimillion-dollar awards.
The first named awards are $480,000 for NPR’s Machine Gods, a video podcast, and $250,000 for NPR’s Power and Influence team. Those grants are concrete commitments; they do not show that the broader $10 million target has been paid out or determine how the rest will be allocated.
Kevin Roose, a Machine Gods co-host, told Semafor that Tarbell’s support helped the project get started and reduced the risk of launching independently. That illustrates what institutional funding can do: give a newsroom the resources to build a reporting project, rather than support only a single story.
Tarbell says its grants can support reporting on frontier-company accountability, policy, AI incidents and labor. Executive Director Cillian Crosson told Semafor, “We think that having resources at newsrooms is an important part of ensuring strong coverage.” The case for funding is persuasive. But resources are not proof that coverage will be more useful or representative; readers still have to judge the journalism and the choices that shaped what got covered.
A firewall needs more than a policy statement
Tarbell says donors do not participate in grant selection or editorial decisions, and that recipient newsrooms retain editorial control. Semafor reported that it found no instances of Tarbell directly intervening in stories. That finding is relevant, but it is not a comprehensive outside review of the selection process or of how the firewall works across the program.
The nonprofit’s funding model also deserves scrutiny. Tarbell’s ethics policy says that Coefficient Giving supplied most of its funding in 2025. Its public donor list names other supporters, including Longview Philanthropy and the Survival and Flourishing Fund. A concentrated funding base does not establish that a donor influenced a grant or a story. It does make clear, however, why readers need to know who funds the funder and how awards are chosen.
Some AI-industry figures and political conservatives, Semafor reported, argue that Tarbell’s funding could encourage coverage that overstates safety concerns. That criticism is not evidence that Tarbell directed particular journalism. The strongest defense is that newsrooms retain editorial control and that reporting on AI’s effects needs resources. Both points matter. Still, a firewall is more credible when the public can inspect the process, not just read the organization’s description of its rules.
Tarbell says it weighs potential impact, audience reach, feasibility and viewpoint diversity. Publishing the reasons for individual awards would let readers compare those stated criteria with the actual choices. A funder can influence which projects have resources without dictating a sentence; that is a structural concern, not an allegation of interference.
Free newsletter
Get the morning briefing
Start each day with the AI stories that matter and why — a short, free email from our newsroom.
Readers should be able to follow the money
Tarbell should maintain a public record of grants awarded and paid, recipients, amounts, project purposes and selection rationales. It should also explain how it expects to distribute the $10 million across outlets, subjects and formats, then update the public as that plan changes. The target conveys scale; it does not reveal priorities.
Disclosure rules need the same clarity. Tarbell requires grantees to disclose its support, but its program rules also give Tarbell advance approval over the wording and format of those notices. That is not evidence that the center controls the reporting. It does mean the funder has a say in how readers are told about its role.
A better safeguard would let each newsroom choose the specific language and placement of its disclosure, subject to a clear minimum standard that identifies the funding and what it supports. Newsroom final say over those details would be a proposed policy change, not the current arrangement. Tarbell should explain how it would resolve disagreements under its existing approval rule rather than leave that question unanswered.
Finally, an independent review of the grant process and editorial safeguards, with public findings, would test whether the written firewall matches practice. That would not certify every story; readers should assess each report on its evidence. It would give them a basis to assess the institution funding it.
Tarbell’s grants could help sustain reporting that workers, users and policymakers need. Tarbell controls the grant rules, and recipient newsrooms control their journalism. To earn public trust, the center should make its allocations traceable, publish the safeguards’ independent review and give newsrooms final say over how they disclose the support.
