Download this commentary (PDF)
In brief...
National leaders recently convened to chart a strategy for U.S. leadership in AI for education and to identify the essentials for building a globally competitive, AI-ready workforce. Many promising ideas were offered, but translating these ideas into federal policy likely will require broader regulatory reform.
On March 11, the Special Competitive Studies Project hosted an AI+Education Summit to address how America’s schools and workplaces can prepare students and workers for an AI-transformed economy. The daylong conference brought together Deputy Secretary of Labor Keith Sonderling, Senators Mark Warner (D-VA) and Mike Rounds (R-SD), former Governors Gina Raimondo (D-RI) and Eric Holcomb (R-IN), NVIDIA co-founder Chris Malachowsky, and education practitioners. The consensus was that the competitive edge of AI lies not in the technology itself, but in developing people capable of deploying it. Senator Warner projected that unemployment among recent college graduates could rise from 9% to 35% within two and a half years. Malachowsky pointed to the supply-demand mismatch driving that risk, observing that only 1% of job seekers currently demonstrate working knowledge of AI while 7% of job postings require it.
Speakers emphasized what LinkedIn’s Catlin O’Neill called the “Five C’s” of curiosity, creativity, courage, communication, and collaboration, over narrow technical fluency. Alex Kotran of the AI Education Project reinforced this, noting that at a major financial services firm undergoing AI transformation, the most productive employees were not the technical staff but those with strong interpersonal skills. Dr. Tasha Arnold of Alpha Schools: AI-powered private school, presented a model in which AI-driven instruction covers core academics in two hours, freeing the remaining six hours for mentorship, entrepreneurship, and life skills, arguing that time in a classroom is not the same as learning. Former Governor Holcomb argued that the future of AI, like past technological upheavals, is inherently unpredictable. His great-grandfather was a blacksmith who could never have imagined his son becoming an aeronautical engineer. But Holcomb stressed that unpredictability is not a reason for inaction; moments of crisis are when real reform happens, and this is that moment. Senator Rounds connected AI readiness to national defense, noting that the U.S. relies on AI across all five warfighting domains: air, land, sea, space, and cyberspace. On the battlefield in Ukraine, he observed that both sides are iterating on their AI capabilities every two weeks, a pace that no traditional training pipeline is built to match. Malachowsky argued the point should be reflected in the language itself: rename “Artificial Intelligence” to “Augmented Intelligence” to keep humans central to the equation.
From Consensus to Implementation: The Regulatory Dimension
The summit’s policy prescriptions focused on expanding apprenticeships, tying funding to outcomes, and, as Senator Rounds advocated, returning most education decisions to the states, on the grounds that local educators and parents are better positioned to adapt than federal agencies competing to award grants. Yet translating these ideas into federal policy likely will require broader regulatory reform.
As Deputy Labor Secretary Sonderling observed, labor laws from the 1930s through the 1960s, governing hiring, wages, performance reviews, and termination, now apply to decisions increasingly made or informed by AI. Employers, developers, and workers, he argued, all need clearer legal obligations for how these tools are built, purchased, and used.
Indeed, the AI tools described by the conference speakers are already being deployed within legal frameworks not built for them. Alpha Schools’ AI tutor continuously adapts to individual students, ingesting behavioral and performance data at a scale that the Family Educational Rights and Privacy Act (FERPA), a statute designed for paper records in 1974, was never built to govern. As the National Education Association has noted, FERPA’s last regulatory updates predate the widespread use of technology in learning environments, leaving schools to interpret a pre-digital law for AI-era data practices. Notably, multiple states have enacted student data privacy statutes that go beyond FERPA’s requirements. Turning to labor law, when AI systems assess candidates in hiring pipelines, they operate under employment discrimination statutes that predate algorithmic decision-making. These legal regimes clearly need a legislative rethink, or else they risk becoming obsolete at best, and inimical to productive uses of AI in schools and workplaces.
There are other implementation challenges. Most of the proposed policy prescriptions involve subsidies to achieve ends that do not seem to have been completely thought through. For example, Sonderling outlined the Department of Labor’s (DOL) push to condition federal workforce funding on AI literacy requirements, but what constitutes “AI literacy” sufficient to satisfy a federal funding condition? Or consider the proposed AI Workforce Training Act (H.R. 7576), introduced in February 2026, which would provide a 30% tax credit for employer AI training expenses. The problem, as Governor Raimondo observed, is that many retraining programs, including those she funded as governor and as Commerce Secretary, delivered mediocre results because they merely funded attendance. Instead, she called for outcome-based funding, noting that the Investing in Tomorrow's Workforce Act (S. 3877) moves in that direction, authorizing competitive DOL grants for AI workforce training pilots. What those outcome metrics should look like remains an open question, though existing federal workforce frameworks under the Workforce Innovation and Opportunity Act offer starting points: job placement rates, credential attainment, and wage gains. But designing these conditions well requires answering questions the summit did not address. What compliance costs would these requirements impose on smaller institutions? When outcome metrics determine which programs survive, who defines those metrics, and how do we guard against perverse incentives that disadvantage programs serving the hardest-to-reach populations?
Regulatory Snapshot
These implementation challenges exist against a broader backdrop: there is currently no coherent federal regulatory strategy on AI in education and the workforce. In the absence of one, states have acted. California finalized employment regulations on Automated Decision Systems effective October 2025, and in May of 2024, Colorado enacted the Artificial Intelligence Act.
Meanwhile, the Trump administration’s March 2026 National Policy Framework for AI pushes to preempt state AI laws while calling for “non-regulatory methods” at the federal level. That tension is already generating political friction from opposite ends of the spectrum: Governor DeSantis has pushed data center regulations in Florida, while Senator Sanders (I-VT) and Representative Ocasio-Cortez (D-NY) have introduced a national moratorium on new AI data center construction.
Senator Warner drew a direct parallel to Congress’s failure to act on social media, where bipartisan intent persisted for years but ultimately produced nothing. The resulting consequences for young people’s mental health are measurable and, as Gallup data cited by Kotran suggests, American anxiety about AI continues to rise each year, particularly among those with less exposure to the technology. This dynamic highlights how poorly designed policy risks fueling the very fears this summit sought to address.
While the summit sounded a note of warning regarding America’s preparedness for an AI-driven world, AI also presents perhaps the greatest opportunity for economic growth since the advent of the personal computer in the 1980s and 1990s. As Senator Warner cautioned, the U.S. cannot use competition with China as a reason to avoid putting structure in place. To realize this potential, effective policy must engage seriously with the known deficiencies of training programs and outdated legal frameworks. New funding conditions, outcome metrics, and literacy mandates must be designed to produce net benefits, not merely signal action, and outdated legal frameworks must not be allowed to quietly undermine the goals this consensus seeks to achieve.