ANA's AI3 Action Institute Puts $3.5M Behind One Winner — and Gives Indian Country 28 Days to Apply

Tribal governments, Native nonprofits, and urban Indian organizations have until August 27, 2026 to compete for a single $3.5 million cooperative agreement to build a national artificial intelligence institute for Indian Country, posted July 30 on Grants.gov as HHS-2026-ACF-ANA-NAI-0035.

That is the whole program. One award. One winner. Twenty-eight days from posting to close, over an August that includes most tribal governments' slowest administrative month. For a community-based organization that has spent the last eighteen months watching federal AI dollars flow to national labs, university consortia, and defense primes, this is the first named, competable notice of funding opportunity where the eligibility list starts with federally recognized tribes and ends without a single research university requirement.

A $3.5 Million Ceiling, a $2.5 Million Floor, and One Expected Award

The AI3 Action Institute listing on Grants.gov is unusually spare for a program with this much conceptual ambition. The Administration for Native Americans — the ACF operating division that runs Native language preservation, environmental regulatory enhancement, and the economic development grants most tribal nonprofits know well — is putting the full $3.5 million behind a single recipient under Assistance Listing 93.612, Native American Programs.

The numbers that matter:

• Award ceiling: $3,500,000
• Award floor: $2,500,000
• Expected number of awards: 1
• Total program funding: $3,500,000
• Funding instrument: Cooperative agreement, not a grant
• Cost sharing or matching: Required
• Posted: July 30, 2026 · Closes: August 27, 2026, 11:59 p.m. ET

The floor is the tell. A $2.5 million minimum means ANA is not entertaining a $400,000 pilot from a single tribe's IT department. The agency has decided in advance that the thing it wants built cannot be built small, and it has structured the competition so that only applicants who can credibly deploy $2.5 million in year one should bother assembling a package.

The cooperative agreement designation matters just as much. Unlike a standard grant, a cooperative agreement carries substantial federal programmatic involvement — ANA staff will sit inside the work plan, not just review the reports. Applicants who write as though they will be left alone for three years will read as though they have not run one of these before. The program contact is Carmelia Strickland (carmelia.strickland@acf.hhs.gov), and in a single-award competition with a four-week window, that inbox is the highest-leverage address in the process.

One scheduling note worth flagging: ANA's own forecast, published in May, projected a June 20 posting and a July 21 due date. The NOFO landed six weeks late with a compressed window. Organizations that built their FY26 calendars off the forecast and stopped checking are the ones most likely to miss this entirely.

Why ANA Built an Institute Instead of Spreading the Money Across 574 Nations

The program description is direct about the intermediary model. The successful applicant "will establish the AI3 Action Institute, which will monitor the pace of AI change, identify responsible and practical opportunities to apply this technology to socioeconomic domains in Native communities, and support its thoughtful integration across Indian Country."

Four functions sit underneath that sentence: reduce the AI learning curve, identify and deploy the best available technologies, optimize the benefits of adoption, and act as a knowledge manager accumulating lessons learned. ANA's framing in the Federal Register was blunter — the institute exists so that Native communities are "firmly in the driver's seat, shaping and leading their own technological futures" rather than absorbing tools designed elsewhere.

Read against ANA's other FY2026 activity, the design logic gets clearer. The agency is simultaneously standing up EAGLE, which replaces the long-running SEDS and SEDS-AK programs with community-driven economic development awards, and NCNTTA, a $3.1 million-per-year cooperative agreement consolidating four previously separate training and technical assistance contracts. AI3 sits in the second lane. It is TTA infrastructure, not direct services — a hub whose output is capability transferred to other organizations, most of which will never be ANA grantees themselves.

That is also the single largest strategic risk in the application. A proposal that describes what the institute will learn, publish, and convene will lose to one that names the tribes, villages, and urban Indian organizations already committed to receiving something concrete in year one.

The Public Comment File Is the Closest Thing to a Scoring Rubric You Will Get

ANA proposed the AI3 policies for comment on March 20, 2026, and issued them final on July 28 — two days before the NOFO posted. That final notice is the most useful unclassified document in this competition, because it records what stakeholders objected to and how the agency answered.

Commenters raised four concerns: opportunity costs, Indigenous data sovereignty, centralized program design, and Tribal readiness. Several questioned whether $3.5 million aimed at emerging technology was the right use of scarce ANA dollars at all, and whether the breadth of issues assigned to a single institute was realistic.

ANA's response was that the institute will "complement — not replace — broader social and economic development efforts" and will be implemented in a way that reflects "ongoing Native input." The final notice also commits the institute to supporting "informed community decision-making" and to addressing "the ethical, practical, and governance considerations associated with emerging technologies."

Translate that into application terms and you get a de facto rubric. Every one of those four objections should be answered explicitly in the narrative, with structure rather than assurance:

• Data sovereignty needs a governance instrument — a tribal data agreement template, an IRB-equivalent review body, a stated position on whether community data ever leaves tribal control for model training.
• Centralization needs a distribution mechanism — regional nodes, subawards, a rotating advisory council seated by tribes rather than appointed by the recipient.
• Readiness needs a tiered on-ramp that assumes some partner communities have limited broadband, not a curriculum that assumes cloud access.
• Opportunity cost needs a measurable socioeconomic outcome in the logic model, not a count of trainings delivered.

The Match Requirement Is the Quiet Filter

ANA programs carry a 20 percent non-federal cost share, with a waiver process available for part or all of the match in defined circumstances. Applicants should confirm the calculation basis in the full announcement, but at 20 percent of total approved project costs, a $3.5 million federal share implies roughly $875,000 in non-federal contribution. Third-party in-kind — donated staff time, tribal facilities, partner-contributed equipment — is typically allowable and is how most ANA recipients actually satisfy it.

Eligibility is broad on the Native side and closed on the commercial side. Federally recognized tribes, incorporated non-federally recognized and state-recognized tribes, tribal consortia, nonprofit Native and urban Indian organizations, Native CDFIs, Alaska Native villages, and tribal colleges and universities can all apply. For-profit organizations cannot — 42 U.S.C. 2991b forecloses it. A tribally chartered business arm can subcontract into the project, but it cannot hold the award.

If You Are Not Going to Win It, You Still Have a Play

For most community-based organizations reading this, the honest assessment is that a four-week window is not enough runway to assemble a competitive $2.5 million-floor institute proposal from scratch. That does not make the deadline irrelevant.

Single-award institutes run on partners. The applicants who are already drafting — likely a tribal college consortium, a national Native nonprofit, or a Native CDFI with existing technical assistance infrastructure — need exactly three things they cannot manufacture internally: named community use cases, letters of commitment from organizations that will actually receive services, and credible reach into populations the applicant does not already serve. An urban Indian health organization with a scheduling backlog, a village corporation with a permitting bottleneck, a tribal Head Start with a caseload documentation problem — each of those is a concrete year-one deliverable a lead applicant would rather cite than invent.

A one-page memo naming your use case, your data constraints, and your willingness to sit on an advisory body, sent this week to the two or three organizations most likely to lead, is a realistic move with a real payoff. Downstream, this institute will be the entity distributing training cohorts, tool pilots, and subawards across Indian Country for years. Being in the original application is the cheapest seat you will ever get at that table.

Federal AI Money Is Shifting Toward Distribution, Not Just Models

AI3 is not an isolated gesture. The same redirect is visible across the federal portfolio — the June 2 AI Executive Order quietly rewired existing grant programs toward AI capability rather than creating new ones, a pattern that has repeatedly caught community-based applicants off guard because the money arrives inside programs they already know rather than under new names.

That is the pattern worth internalizing. Agencies are increasingly funding the adoption layer — institutes, hubs, technical assistance intermediaries — because the models themselves are already commodity infrastructure. For organizations serving rural, tribal, and faith communities, this is a more accessible funding posture than the research-grant era it replaces, and it is arriving at the same moment that core anti-poverty infrastructure like the Community Services Block Grant is being dismantled. Capability funding is growing while operating funding shrinks. Organizations that can convert the former into the latter will be the ones still standing in FY2028.

What to Do Before August 27

If you intend to lead: pull the full announcement from Grants.gov today, email Carmelia Strickland with your governance question this week, and build the narrative around the four comment-file objections rather than around your capabilities statement. If you intend to partner: send the one-page use-case memo now, while lead applicants are still assembling attachments.

Either way, start by mapping what else is open in this lane. Search active tribal and Native AI funding opportunities on Granted — AI3 is one of several ANA and cross-agency notices live right now, and the ones with longer runways are where a first-time applicant has better odds than a twenty-eight-day sprint.