A New Compact — The Compact Institute

Introduction

Today we’re announcing The Compact Institute, a new project on AI, technological transitions, and governance.

Every major technological transition is a choice point, or a moment when political and economic power are reorganized. We can shape this moment through new tools, partnerships, policy mechanisms, and institutional design. We believe this restructuring should not only mitigate the risks of new technologies, but also harness new capabilities on the frontier to enable better futures.

Today’s technologies, from AI and energy to biotechnology and space, are already massively changing our markets, systems, and institutions. In the postwar period, institutions like the NSF, NIH, and DARPA operationalized a technology-society compact that transformed innovation. We now have the opportunity to form a new compact between the technological frontier and the public—and to build the institutions necessary to realize that vision.

Why past compacts no longer work

In the mid-1930s, private electric utility companies had little financial incentive to run power lines to sparsely populated areas. Forty years after the construction of the first power plants, 90% of urban homes had power, but only 10% of rural ones did. Rather than subsidize private utilities or wait for markets to adapt, FDR and the New Deal coalition created the Public Utility Holding Company Act, Rural Electrification Administration, the Tennessee Valley Authority, and community-owned financing structures.1 By 1950, the rural/urban electricity gap had narrowed to 10%. Together, these institutional reforms formed a compact that redefined the industry’s relationship to the public: once a technology became infrastructure, access to it was a public obligation that could be regulated by new ownership rights.

A decade after the New Deal, we revised the technology-society compact when the times changed and demanded it. With the war’s end in sight in November 1944, President Roosevelt asked Vannevar Bush, who had directed the wartime research effort, how that same scientific capacity could be carried into peacetime. Bush answered in his landmark report, Science, the Endless Frontier. He proposed a bargain: the government would fund basic research while scientists and universities would retain full autonomy over their research direction, and in return that free inquiry would pay off in national security, public health, and economic growth. Institutions like the National Science Foundation operationalized this compact.

We’ve been living within this compact for eighty years, even as the nature of technology has fundamentally changed in several ways.

Different institutions produce breakthrough technologies. In 1964, the federal government funded 67% of all American R\&D; by 2022, it funded 18% while private industry funded 75%. This relationship has inverted. For emerging technologies, the gap is even starker. More than 90 percent of notable AI models come from industry rather than academia, and Amazon, Google, Meta, and Microsoft have announced roughly $700 billion in combined capital expenditure for 2026, several times what the federal government spends on R\&D in a year. Bush’s compact assumed that well-funded universities and national laboratories would work at the frontier. For AI, it structurally cannot, because the capital requirements2 exceed any budget Congress could plausibly appropriate.
The private firms building transformational technology operate at a planetary scale. They are reshaping structures of sovereignty rather than being contained by existing governance units, like nation-states. OpenAI runs the OpenAI for Countries program, whose product is sovereign AI capacity for national governments. Its first deployment, Stargate UAE, is a 1-gigawatt cluster in Abu Dhabi. A private firm is now a party to arrangements that constitute another state’s technological capacity.
Adoption today happens much faster. Whereas physical infrastructure like electrification took decades to reach half of American households, recent general-purpose AI tools, backed by vast amounts of capital, have reached comparable penetration in just a few years. In 2026, nearly half of American adults reported having used an AI chatbot (Pew Research). Our institutions are calibrated to multi-year cycles of committee hearings, markup, and notice-and-comment rulemaking, and weren’t built for this pace3.

We need a new technology compact fit for this age. The Compact Institute is our contribution to that effort.

What a new compact looks like

We think a new compact has to get a few principles right.

The public should have real means to set the direction of new technology (public steering), and the resulting gains should be legible: visible in prices, wages, services, infrastructure, and capabilities that people can name (public benefit).
As technological capabilities develop rapidly, policies and tools to implement them should be developed with appropriate feedback loops, monitoring, and sunset provisions, so governance can learn closer to the speed of what it governs (adaptive governance).
Technology development today tends toward centralization, with a handful of firms, models, and standards becoming the default. A plurality of firms, approaches, and institutions is a property of healthy technological ecosystems, surfacing failure modes and better designs. Plurality is also an important check on the concentration of power. Even an actor who may be well-intentioned , like a firm or a government agency, becomes a risk once it controls a critical technology exclusively.

What we’re working on

Building toward this compact, we’ve organized our initial work around four mechanisms we think are underused relative to their potential impact.

Creating tools. As the world becomes more complex, we need new ways to understand it.

Mapping AI: a public map of who shapes U.S. AI policy, what they believe, and how they are connected. We’ve mapped 2,041 people and organizations and 2,978 relationships, covering affiliations, funding, policy positions, AI timelines, risk views, and sources. | Launch | Marketplace interview | Research insights | Methodology
Power Concentration Mapping (ongoing): Every policy intervention implicitly assumes a threat model. We want to systematically map these threat models–their assumptions, causal pathways, and the measurement systems needed to track them. | Research concept note | Prototype power concentration index | Radiant Example forecasting map with Metaculus

Building partnerships and coalitions. Frontier technology governance requires coordination, but relevant expertise is often scattered.

Lessons ahead of 2028 (ongoing): We plan to interview losing primary candidates about their theory of change and diagnosis of failure points to inform strategy for the 2028 presidential election. We suspect losing candidates will be more candid about their experience, especially about lobbying interests that approached their campaigns. | Concept Note
Fieldbridging (ongoing): We want to build strategic relationships with pro-democracy groups, corporate and constitutional law scholars, and other fields that are motivated to prevent concentration of power. By illustrating the risks of AI-enabled power concentration through pathways like democratic backsliding, coups, and capital accumulation, we hope to increase support for robust technical governance and policy interventions.

Experimentation and gathering evidence. Effective governance depends on testing interventions before scaling. However, methods like RCTs and quasi-experimental designs require comparisons and evidence that don’t exist for current policy proposals. We want to build a suite of alternative methods, including several that use AI itself. We’re also exploring new mechanisms to share and produce data.

Measuring the AI economy (ongoing): A pilot proposal for structured transparency between frontier labs and government statistical agencies, to enable real-time data sharing about AI’s impacts on the economy. | Working paper

Reforming and designing new institutions. Existing institutional forms, such as standard-setting bodies, scientific assessment panels, and shareholder trusts, each carry their own functions and relative advantages. Future modes of organization should be designed to fill the capability gaps, such as speed, verifiability, and technical oversight, that our existing institutions miss.

Whose Company Is It? (ongoing): Today’s corporate structures don’t hold AI megacorps accountable for their effects on the public. We’re exploring new institutional forms, such as an enforceable public fiduciary, that bind firms to account for the public interest through both internal governance and verifiable external checks.| Working paper: Part 1 | Research concept note
Functionalism as institutional design (ongoing): Different institutional forms are suited to different problem spaces (see: FROs for building shared scientific research infrastructure; standards bodies for decentralized coordination). Applied to AI, this points to concrete gaps–for instance, an IAEA-style scientific body for verifying compute and training claims, a NTSB-style bureau for investigating AI incidents across companies, and a FRO-style entity for funding shared safety infrastructure like evals and interpretability tools. This project would build a working catalog of problems matched to institutional types and propose new forms where necessary.

Conclusion

Every technological transition offers a window to build better institutions, infrastructure, and futures. Whether or not we take advantage of this opportunity is up to us.

In forming a new compact, we borrow deliberately from ways of working that developed outside of traditional policy work: from startups, the philosophy of iterating quickly and putting work in front of users; from organizing, the understanding that durable change requires both broad coalitions and disruptive demonstration; and from engineering, the instinct to build tools and design rigorous experiments.

We learn from newer organizations connecting policy issues to implementation and state capacity, including IFP, the Niskanen Center, and Code for America; from evidence-based policy research from orgs like RAND, Brookings, and the Urban Institute; and from the long-horizon thinking of Forethought and the Berggruen Institute. More traditional DC think tanks, such as the Cato Institute, Heritage Foundation, Center for American Progress, and the World Resources Institute offer different models for developing policy agendas and shaping public discourse over time. We also draw from incubators built for innovation, such as the Bell Labs, DARPA, the MIT Media Lab, and Y Combinator, alongside the coalition-building traditions of the Industrial Areas Foundation and People’s Action. We’ve started collecting a library of the work that has shaped our theory of change so far.

This document is a living charter, and we would like to hear what’s missing and opportunities to work together. Please contact: info@compactinstitute.org