$10 Million, 16 Nonprofits, and OpenAI Engineers: Inside the Private Fund That Is Quietly Building the AI Safety Net

While Congress debates whether AI will destroy jobs or create them, a private fund has been quietly writing checks to the organizations building the answer. In March 2026, the GitLab Foundation announced its largest cohort yet for the AI for Economic Opportunity Fund — 16 organizations selected from more than 800 applications, each receiving $250,000 in catalytic funding plus something money alone cannot buy: six months of dedicated technical support from OpenAI engineers.

Since launching in 2023, the fund has invested nearly $10 million across roughly 50 organizations working at the intersection of artificial intelligence and economic mobility. As Granted News reported, the latest cohort represents a significant scaling of the program's ambition — and its partners. The GitLab Foundation now works alongside OpenAI, the Annie E. Casey Foundation, and Ballmer Group to create what has become one of the most active philanthropic vehicles deploying capital where AI meets poverty.

The numbers the fund projects are striking: $1,735 in per-person annual earnings increases, $52,035 in per-person lifetime earnings increases, 27,478 total beneficiaries, and an aggregate $1.43 billion in lifetime earnings impact. Those projections are, of course, projections. But the specificity of the funded projects — and the engineering resources backing them — suggests this is not a grant program that measures success in reports filed.

What $250,000 Plus Engineers Actually Buys

Most grant programs hand over a check and wish you luck. The AI for Economic Opportunity Fund operates differently, and the difference matters.

Each grantee receives $250,000 in unrestricted funding — enough to build a prototype, hire a technical lead, or run a pilot. But the more unusual component is the technical support: six months of cohort-based mentorship from OpenAI Academy engineers, direct access to OpenAI's API with credits included, and structured collaboration with the other 15 organizations in the cohort.

For a nonprofit trying to build an AI-powered benefits navigator or a career matching system, the gap between "we have an idea" and "we have a working product" is almost never money alone. It is technical capacity — the ability to architect a system that works at scale, handle the edge cases that break prototypes, and build on foundation models without burning through API credits experimenting. The OpenAI engineering partnership closes that gap in a way that a $250,000 check by itself never could.

After the initial six-month build phase, grantees enter a demonstration period where high-performing projects become eligible for additional funding from the partner philanthropies. This staged model — seed, build, demonstrate, scale — mirrors venture capital more than traditional philanthropy, and that is intentional. The fund's theory of change is that the organizations best positioned to use AI for economic mobility are the ones already working closest to the problem, and what they need is not a five-year research grant but rapid technical capacity and a runway to prove their approach works.

The 16 Projects: What They Build and Who They Serve

The funded organizations are not building AI for its own sake. Each project targets a specific failure point in the economic mobility system where AI can remove friction, expand access, or make existing services work better for more people.

Benefits navigation emerges as the strongest theme. Moms First is building a navigator that connects low-income parents to benefits they are already eligible for but are not receiving — projected to reach 75,000 families and deliver $3,000 to $8,000 in annual benefits per household. Scholar Fund is simplifying benefits applications. U.S. Digital Response is building language access tools for Maryland's Medicaid, SNAP, WIC, and TANF programs, addressing the barrier that keeps non-English-speaking residents from accessing programs designed to serve them.

The connection to federal grant policy is direct. As the anti-fraud task force executive order increases verification requirements for these same benefit programs, organizations that help eligible people navigate the system become more important, not less. Complexity is the enemy of access, and AI-powered navigation tools reduce complexity.

Career matching and workforce development is the second major cluster. Per Scholas, a national tech training nonprofit, is using AI to scale its training capacity to 15,000+ learners with projected wages 2.5 times pre-training levels. SkillUp Coalition is building AI-guided career navigation for 500,000 job seekers. The MIT Media Lab is developing a proactive reskilling identification system that flags workers at risk of displacement before their jobs disappear — targeting 18 million at-risk workers and aiming to prevent 15-25% wage losses.

Career Path Services is building AI navigation tools with a targeted 10% improvement in job placement rates and 20% increase in training enrollment. Colorado Thrives is constructing an AI-powered career marketplace aiming for 1,000 job placements. Roadtrip Nation is deploying AI career exploration guidance for 3,200+ learners. Moses/Weitzman Health System is creating longitudinal career mapping with mentorship matching for 500,000+ workers.

Global economic development rounds out the portfolio. The Development Innovation Lab at the University of Chicago is building AI-enhanced weather forecasts for farmers, targeting 100 million farmers in low- and middle-income countries — a project where better forecasts directly translate to better planting decisions, reduced crop losses, and higher incomes. Accion is deploying AI-powered business intelligence for pharmacy professionals in Kenya, projecting a 21-33% income boost for 16,000 professionals. The National Domestic Workers Alliance is building training and legal support tools to scale services to 220,000 domestic workers.

NASWA/CESER, the national association of state workforce agencies, is building apprenticeship connection tools — projected to connect 15,000+ students with apprenticeships that average $80,000 in starting salary. And the Community Economic Defense Project is creating AI-powered legal guidance for renters facing eviction, with the potential to reach 3.6 million households.

The Foundation for California Community Colleges brings scale: AI-driven pathway identification for 2.1 million students, targeting $12,000 to $20,000 in individual earnings increases.

Five Patterns Worth Watching

The GitLab Foundation identified five emerging trends among this cohort that signal where AI-for-good philanthropy is heading — and where the next round of funding opportunities will likely emerge.

Predictive policy infrastructure. Several grantees are not building standalone tools but embedding AI directly into government systems. U.S. Digital Response's work with Maryland state benefits is a template: rather than building a parallel system, they are adding an intelligence layer on top of existing infrastructure. This approach is cheaper, faster to deploy, and more likely to survive changes in administration because it improves what already exists rather than competing with it.

Scalable human services. The hardest problem in social services is not knowing what to do — it is doing it for enough people. Case management, career counseling, legal guidance, and benefits navigation all work well when delivered by skilled humans. They all break when demand exceeds capacity. AI that augments rather than replaces human services workers — handling intake, triaging cases, generating first drafts of applications — allows the same workforce to serve dramatically more people.

Benefits navigation as a category. The number of grantees working on benefits access suggests the philanthropic community has identified a clear failure point in the American safety net. Eligible families do not receive benefits they qualify for because the application processes are too complex, the information is scattered across agencies, and the systems are not designed for the people who need them most. AI can make these systems legible.

Labor market intelligence. The MIT Media Lab's proactive reskilling project and NASWA's apprenticeship matching tool both represent a shift from reactive job search to predictive career navigation. Instead of helping people find jobs after they lose them, these tools aim to identify displacement risk and reskilling pathways before job loss occurs.

Global reach from domestic base. The inclusion of the Development Innovation Lab's farmer weather forecast project and Accion's Kenya pharmacy intelligence platform signals that this fund does not see AI economic opportunity as a domestic-only proposition. The same models and approaches that work in American workforce development can be adapted for global markets where the stakes — and the potential impact per dollar — are even higher.

What This Means for the Grant Landscape

The AI for Economic Opportunity Fund is not the only private fund operating in this space, but it is among the most operationally sophisticated. The combination of unrestricted cash, dedicated engineering support, cohort-based learning, and staged follow-on funding creates a pipeline that federal grant programs — with their 12-to-18-month award timelines, rigid reporting requirements, and siloed program offices — cannot match for speed or flexibility.

That does not make federal funding irrelevant. The scale of federal programs dwarfs what any private fund can deploy. But as federal grant uncertainty continues — between DOGE terminations, agency restructuring, and the new anti-fraud enforcement apparatus — private philanthropic funds like this one become increasingly important as proving grounds for innovations that federal programs may eventually adopt.

For nonprofits considering an application to the next cycle, the signal is clear: the fund wants organizations already working at the front lines of economic mobility, not AI startups looking for their first customer. The technical capacity comes from OpenAI. What the fund cannot supply is deep understanding of the problem you are solving and the relationships with the communities you serve.

If you are building something at the intersection of AI and economic access — whether it is benefits navigation, career development, legal aid, or workforce training — the fact that 800 organizations applied for 16 spots tells you both how competitive this is and how much demand exists for this kind of support. Tracking opportunities like this, and building the strongest possible application when they open, is the kind of strategic work that Granted helps organizations do systematically rather than accidentally.