PUBLIC AI INFRASTRUCTURE / NEW YORK STATE
EMPIRE AI
WHO CONTROLS PUBLIC-INTEREST ARTIFICIAL INTELLIGENCE?
New York is investing in a new model of artificial intelligence infrastructure, bringing together public resources, academic institutions, private technology, and advanced computing capabilities.
The opportunity is significant.
So are the questions.
Who governs the infrastructure? Who receives access? Which communities benefit? And how can public investment produce genuinely public-interest AI?
PUBLIC INVESTMENT IS NOT THE SAME AS PUBLIC CONTROL.
SECTION 01
WHAT IS EMPIRE AI?
Empire AI is a New York-based initiative intended to expand advanced artificial intelligence research capability through shared computing infrastructure and institutional collaboration. Its own published framing is direct: AI research for the public good, placing world-class computing power in the hands of New York researchers to accelerate discovery, solve real-world problems, and advance the public good.
The initiative emerged from New York State's push to keep advanced AI research capacity inside public and academic institutions rather than leaving frontier-scale computing exclusively to private laboratories. Announced by the Governor as a flagship public investment and developed with SUNY and partner institutions, it pairs state resources with university participation and private-sector technology contributions.
Its published research pages already show the intended division of labor in practice: Columbia University researchers measuring New York City traffic volume to support transportation and public-safety decisions; University at Buffalo teams analyzing billions of molecular interactions to accelerate drug discovery; University of Rochester researchers simulating surface-reaction chemistry to improve manufacturing efficiency; a CUNY team predicting hyperlocal flooding to help New York City prioritize emergency response; and NYU researchers building one of the largest public datasets of AI-generated conversations to study how synthetic training data shapes large language models.
This page does not treat any of that as a verdict. It treats it as the starting point for a set of questions every publicly supported AI institution should be prepared to answer.
FOUNDING PURPOSE AND DEVELOPMENT
Documented. Empire AI describes itself as a public-interest computing initiative for academic research; its public-facing materials and research examples are consistent with that framing.
NEW YORK STATE'S ROLE
Partially documented. The Governor's office announced the initiative as a signature public investment, and state entities participate in its structure. Exact appropriation figures, budget lines, and procurement terms have not been verified against primary budget records for this page.
PARTICIPATING INSTITUTIONS
Partially documented. Published research examples attribute work to Columbia University, CUNY, New York University, the University at Buffalo, and the University of Rochester. The full consortium membership roster is maintained by Empire AI and has not been independently verified for this page.
PUBLIC AND PRIVATE FUNDING COMMITMENTS
Not yet verified. Public reporting indicates a combination of state investment, institutional commitments, and private technology contributions. Announced totals, the obligations attached to each contribution, and actual expenditures to date require verification against official budget and procurement records before this page states them as findings.
COMPUTING INFRASTRUCTURE AND CAPABILITIES
Partially documented. Empire AI operates from 4 International Drive, Rye Brook, New York, and announced the launch of its Beta phase in August 2026. Detailed technical specifications of the computing environment have not been verified for this page.
RESEARCH ACCESS AND ELIGIBILITY
Partially documented. The organization's site includes a registered-user access path for researchers. Eligibility criteria, allocation procedures, and any expanded-access arrangements have not been verified for this page.
STATED OBJECTIVES
Documented. Empire AI's stated objectives are to accelerate scientific discovery, solve real-world problems, expand access to computing power typically available only to large corporations, and deliver meaningful benefits for New Yorkers.
Two distinctions frame everything that follows:
Announced funding commitments are not actual expenditures. A pledge is a plan; a budget line is a record. This page distinguishes the two and marks totals as pending verification until official records confirm them.
Proposed capabilities are not operational capabilities. What an initiative says it will make possible and what researchers can do today are different claims, and they age differently.
PRIMARY SOURCES FOR THIS SECTION
Empire AI — official website (who-we-are, research, and news pages)
VERIFIED — reviewed 2026-11-12
www.empireai.edu/ ↗Empire AI — Governance (Board of Directors and leadership team)
VERIFIED — reviewed 2026-11-12
www.empireai.edu/governance/ ↗Governor of New York — official announcements and budget materials
DOCUMENTATION REVIEW IN PROGRESS
www.governor.ny.gov/ ↗State University of New York (SUNY) — system communications
DOCUMENTATION REVIEW IN PROGRESS
www.suny.edu/ ↗Participating universities' official announcements
DOCUMENTATION REVIEW IN PROGRESS
New York State budget and procurement records
DOCUMENTATION REVIEW IN PROGRESS
SECTION 02
WHO PAYS? WHO BENEFITS?
Public resources support Empire AI. The public-interest question is what the public receives in return, on what terms, and who decides. These questions apply to every subsidy model, and none of them is answered by the fact that the investment is well intentioned.
- How much public funding has been committed, through which appropriations, and over what years?
- What private contributions are involved, and what obligations accompany them?
- Who owns or controls the resulting infrastructure, and under what arrangement?
- Who owns the intellectual property generated through supported research — and can the public benefit from it without restriction?
- How are public benefits measured, and by whom?
- What happens if commercial entities profit from publicly supported research?
- Are the financial and institutional arrangements transparent enough for independent verification?
PUBLIC SUBSIDY
Public money lowers the cost of an activity. It does not, by itself, give the public a say in how the activity is governed.
PUBLIC OWNERSHIP
Public ownership of assets or outcomes creates enforceable public claims. It is a distinct legal condition, not a synonym for subsidy.
PUBLIC ACCESS
Access determines who can use the infrastructure and on what terms. Broad eligibility turns capacity into shared capability; narrow eligibility turns it into a gate.
PUBLIC ACCOUNTABILITY
Accountability means the public can examine decisions, contest them, and obtain remedies. It is a governance property, not a funding property.
These four are not interchangeable. An initiative can be publicly subsidized without being publicly owned, publicly accessible without being publicly accountable. Conflating them is how public investment quietly becomes private advantage.
SECTION 03
WHO GOVERNS EMPIRE AI?
Empire AI publishes its governance structure. As of the last review date on this page, its published Board of Directors lists the following members and affiliations:
Tom Secunda
Bloomberg Co-Founder (Chair)
Joshua Brumberg
CUNY
Stephen Dewhurst
University of Rochester
Liz Fine
Principal, Liz Fine Associates LLC
Ian Fisk
Simons Foundation (Observer)
Venu Govindaraju
University at Buffalo
Nina Gray
NYU
Douglas Grose
NY Creates
Kim Hazelwood
Technology executive
Robert Hull
RPI
Howard Morgan
B Capital
Greg Morrisett
Cornell
Girish Nadkarni
Icahn School of Medicine at Mount Sinai
Ram Ramasubramanian
SUNY
Ben Verschueren
NYSTAR (Advisor)
Jeannette Wing
Columbia University
Composition as published by Empire AI and reviewed on the date shown at the top of this page. Affiliations are reproduced as listed on the governance page; titles and roles may change, and this list should be re-verified at each review.
Institutional representation is substantial: CUNY, SUNY, Columbia, Cornell, NYU, RPI, Rochester, and Buffalo are all present, along with NYSTAR in an advisory seat.
Private-sector and philanthropic presence is likewise substantial: the chair is a Bloomberg co-founder, and the published board includes a technology executive, a venture-capital figure, and a foundation observer.
No seat on the published board is labeled for civil rights organizations, community institutions, public-interest technologists, or groups affected by AI deployment. Whether such representation exists elsewhere in the governance structure is an open question, not an established finding.
OPEN QUESTIONS
- Do conflicts-of-interest procedures apply to board members who hold roles in commercial technology or investment firms, and are they published?
- How are research priorities set, and what role do public stakeholders play in that process?
- Are board and committee proceedings transparent to the public, and are minutes or materials published?
- What independent oversight exists beyond the organization's own governance?
- What grievance and appeal mechanisms are available to researchers, affected communities, or members of the public?
An absence of publicly available information is not evidence that a safeguard or mechanism does not exist. Every item above is labeled a question for exactly that reason — and each deserves a documented answer.
SECTION 04
WHO GETS TO BUILD THE FUTURE?
This is the question on which public AI infrastructure most directly meets its promise — or fails it. Computing power is not distributed evenly across the research landscape, and the design of an access regime determines whether public investment narrows that gap or reproduces it.
MAJOR RESEARCH UNIVERSITIES
Well-resourced institutions with dedicated research computing staff can absorb allocation processes that smaller institutions cannot. Does Empire AI's access model account for that asymmetry?
SMALLER ACADEMIC INSTITUTIONS
Can teaching-focused and regional institutions realistically compete for allocations, or do administrative capacity and technical staffing operate as informal filters?
HISTORICALLY BLACK COLLEGES AND UNIVERSITIES AND MINORITY-SERVING INSTITUTIONS
Do HBCUs and MSIs hold consortium membership, dedicated allocation pathways, or support structures — or does participation default to whoever already has infrastructure?
COMMUNITY COLLEGES
Are community colleges — the broadest on-ramp to technical education in New York — within scope at all?
INDEPENDENT RESEARCHERS AND RESEARCHERS WITHOUT INSTITUTIONAL AFFILIATION
Is there any pathway for credentialed independent researchers, or does eligibility presuppose institutional affiliation?
NONPROFIT CIVIC TECHNOLOGY ORGANIZATIONS
Can public-interest organizations building civic tools access supported computing, or is access confined to academic research?
EARLY-STAGE RESEARCHERS AND ENTREPRENEURS
Do students, postdocs, and first-time principal investigators have realistic routes to allocations, or does access concentrate among established labs?
UNDERREPRESENTED COMMUNITIES
Does the access model create conditions under which the communities most affected by AI systems can build, evaluate, and study those systems themselves?
Participation can be limited by formal eligibility, technical requirements, institutional affiliation, funding, administrative capacity, or other documented barriers. Which of these apply to Empire AI's access regime is a factual question that requires its published allocation policies to be verified.
THE FUTURE OF ARTIFICIAL INTELLIGENCE WILL BE SHAPED NOT ONLY BY WHO USES IT, BUT BY WHO HAS THE RESOURCES TO BUILD IT.
No demographic disparity should be attributed to Empire AI without supporting data — and no claim of equitable access should be accepted without it either. The published research examples come from major research universities. What the access regime looks like for everyone else is a question this page holds open.
SECTION 05
PUBLIC INFRASTRUCTURE. PRIVATE TECHNOLOGY.
Publicly supported AI infrastructure runs on a technological supply chain that is overwhelmingly commercial: accelerators, servers, networking, cloud capacity, software frameworks, and data-center operations. Examining that dependency is not an accusation — it is a description of the environment in which every AI institution currently operates.
- Computing hardware: advanced accelerators are manufactured by a small number of firms, and public reporting has associated NVIDIA with Empire AI's computing environment. The precise terms — procurement, donation, or partnership — have not been verified for this page.
- Cloud and data-center operations: large-scale training and serving environments depend on commercial facilities and operational expertise.
- Networking equipment and software platforms: the transport and software layers of the stack are commercial products with their own concentration dynamics.
- AI development frameworks: the dominant toolchains for building and training models are maintained by a handful of companies.
- Commercial research partners: collaboration with private firms is a documented feature of the initiative's model, and its terms matter as much as its existence.
These relationships are not all the same thing, and treating them interchangeably produces bad analysis:
Procurement is a purchase. Donation is a gift with whatever expectations follow it. Research partnership is a working relationship with negotiated terms. Technical dependency is a structural condition. Governance influence would be a different and more serious matter.
Using a vendor's products does not mean the vendor controls the research agenda. This page does not imply otherwise — and the question of whether governance influence exists is answerable only through documented decision-making records.
This analysis connects directly to DSC's SCARCITY REGIME TRIANGLE — the structural dependency of advanced computation on three constrained inputs:
ENERGY
Large-scale computing is an energy question before it is a software question. Who bears the cost, and who bears the externalities?
HARDWARE
Accelerators, servers, and networking equipment concentrate in a small commercial market. Public infrastructure does not escape that concentration by purchasing within it.
SOFTWARE
Frameworks, orchestration layers, and model toolchains shape what research is practical to attempt.
The distinction that matters here is between infrastructure access and infrastructure sovereignty. Access means New York's researchers can use advanced computing. Sovereignty would mean the capability to sustain, govern, and directionally shape that computing without structural dependence on markets the public does not control. Empire AI can deliver the first without delivering the second — and whether it can deliver both is one of the defining questions of public AI infrastructure.
SECTION 06
WHAT PROTECTS THE PUBLIC?
Advanced AI infrastructure needs credible governance even when — especially when — its stated mission is beneficial. Beneficial intent does not substitute for enforceable safeguards, and the absence of demonstrated misconduct is not a substitute for their existence.
- Research ethics: what review processes apply to supported research, and are they published?
- Privacy and data protection: what standards govern the data used in supported research, and how do they interact with New York and federal law?
- Human-subject research oversight: when supported research involves people, what oversight applies?
- Algorithmic discrimination: what evaluation obligations attach to systems whose deployment could affect rights or life chances?
- Biometric and sensitive data: are there category-specific restrictions on the data supported research may process?
- Surveillance-related applications: what limits exist on using public computing resources for surveillance research, and who enforces them?
- Dual-use research: how does the initiative handle research with plausible harmful applications?
- Cybersecurity: what standards protect the infrastructure and the research conducted on it?
- Public-sector AI applications: when state or local agencies draw on supported research, what additional accountability applies?
- Research transparency and independent auditing: can outsiders examine what is being built and how it performs?
For each of these, the operative question is the same: is the protection established, is it institution-specific, or is it unclear?
The risks of AI research generally are not demonstrated misconduct by Empire AI. This page does not conflate the two — and it does not accept the inverse, that a beneficial mission removes the need for safeguards. Both propositions are analytical, and both must stand on evidence.
SECTION 07
WHY NEW YORK MATTERS
New York is one of the few places where the full civic ecosystem for public AI infrastructure exists in a single jurisdiction: major research universities, scientific institutions, technology companies, civic organizations, public agencies, cultural institutions, and one of the most diverse populations in the world.
That combination creates a genuine opportunity: to demonstrate that AI innovation can coexist with civil liberties, public accountability, equitable access, and independent civic research — not as a marketing claim, but as a working institutional model other states and cities could study.
It also raises the stakes of the questions on this page. If a state of New York's scale and diversity cannot produce a public AI model that is demonstrably accountable and accessible, the lesson other jurisdictions draw will be that public AI infrastructure is a subsidy program rather than a democratic institution.
Digital Sovereignty Coalition's own origins are in New York, and its digital civil rights mission was built on the premise that the communities most affected by technology must have a voice in how it is governed. That premise is exactly what makes the questions on this page worth asking well — here, first.
This section describes civic context. It does not imply any formal relationship between DSC and Empire AI; see the independence disclosure at the end of this page.
SECTION 08
CAN PUBLIC AI INFRASTRUCTURE SUPPORT CIVIC AI?
Civic AI, as DSC uses the term, is a public-interest principle: AI capabilities designed to support human autonomy, civil liberties, democratic accountability, and meaningful public participation. It is a testable question whether publicly supported research infrastructure could eventually enable it.
- Independent civic AI research
- Public-interest computational tools
- Privacy-preserving civic services
- Community-centered AI research
- Independent algorithmic evaluation
- Public-interest research collaborations
- Noncommercial AI experimentation
DSC's proposed conceptual civic AI architecture, Civic Machina, is a framework for accountable AI — not an operating platform, and not an Empire AI project. Nothing on this page should be read as connecting the two.
Empire AI presents an opportunity to examine whether advanced computing infrastructure can support a plurality of AI development models, including nonprofit and civic-interest approaches. This possibility should be explored, not presumed.
The principles behind this question are set out in DSC's response framework:
READ THE RIGHT TO CIVIC AISECTION 09
BEYOND THE COMMERCIAL AI MONOCULTURE
Examined through DSC's Plurality of Digital Sovereignty framework, the fundamental issue is not whether a publicly supported research institution can compete with private frontier AI laboratories. It is whether alternative institutions can develop independent capabilities, governance models, research priorities, and public accountability.
These are analytical categories for evaluating institutional models — not claims about Empire AI's formal structure.
Plural institutional models could reduce excessive concentration without undermining innovation. Whether that pluralism emerges depends less on any single institution's success than on whether the access, governance, and accountability conditions that let alternatives take root are deliberately designed. That is the design question beneath this entire page.
SECTION 10
QUESTIONS WORTH ASKING
A standing list of unresolved questions concerning governance, access, public benefit, corporate dependencies, civil rights, and research independence. Each is classified by its current evidentiary status. These classifications describe the state of public documentation — they are not allegations, and they carry no implication of wrongdoing.
Who chairs Empire AI's board, and what is the board's composition?
ANSWERED BY PUBLIC DOCUMENTATION
empireai.edu/governance — board roster and affiliations as published
Last reviewed:
What research is being conducted on Empire AI infrastructure?
ANSWERED BY PUBLIC DOCUMENTATION
empireai.edu — published research examples (Columbia, CUNY, NYU, University at Buffalo, University of Rochester)
Last reviewed:
What is the total committed public investment, and what has actually been expended?
NOT ADDRESSED IN PUBLISHED DOCUMENTATION
Announced figures reported publicly; official budget and procurement records not yet verified for this page
Last reviewed:
Which institutions hold consortium membership?
PARTIALLY DOCUMENTED
Research pages attribute published work to five institutions; the full roster has not been verified
Last reviewed:
Do civil rights organizations, community institutions, or public-interest technologists hold governance roles?
NOT ADDRESSED IN PUBLISHED DOCUMENTATION
No seat on the published board is labeled as such; absence from a published list is not evidence of absence
Last reviewed:
What are the eligibility criteria and allocation procedures for computing access?
NOT ADDRESSED IN PUBLISHED DOCUMENTATION
A registered-user access path exists; its criteria have not been verified
Last reviewed:
What obligations accompany private contributions to the initiative?
NOT ADDRESSED IN PUBLISHED DOCUMENTATION
Terms of private participation not published in materials reviewed for this page
Last reviewed:
Who owns intellectual property generated through supported research?
NOT ADDRESSED IN PUBLISHED DOCUMENTATION
IP arrangements not published in materials reviewed for this page
Last reviewed:
What civil rights, privacy, and audit safeguards apply to supported research?
NOT ADDRESSED IN PUBLISHED DOCUMENTATION
Safeguard documentation not located in materials reviewed for this page
Last reviewed:
How are public benefits measured, and are the measurements published?
NOT ADDRESSED IN PUBLISHED DOCUMENTATION
Benefit metrics not published in materials reviewed for this page
Last reviewed:
Page last reviewed:
SECTION 11
INSTITUTIONAL DEVELOPMENTS AND RESEARCH TRACKER
This section is a standing tracker for future updates. Every entry added here must include its date, its source, and a short editorial note on its significance. Entries are added only when a development is verified against primary documentation.
Funding and appropriations
Governance changes
Research partnerships
Infrastructure developments
Research access policies
Civil rights and privacy safeguards
Public-interest research outcomes
No entries have been added yet. This page does not generate unsupported news; the tracker will grow only as developments are independently verified.
SECTION 12
INNOVATION AND ACCOUNTABILITY ARE NOT OPPOSITES
Digital Sovereignty Coalition recognizes the potential value of publicly supported artificial intelligence research infrastructure.
Expanding access to advanced computing may create opportunities for scientific discovery, institutional independence, and new models of public-interest innovation.
But public investment alone does not guarantee democratic accountability, equitable participation, technological independence, or meaningful public benefit.
Those outcomes depend on governance, transparency, access, enforceable safeguards, and independent scrutiny.
The objective is not to obstruct AI innovation.
It is to ensure that innovation serves society without requiring society to surrender its autonomy.
THE QUESTION IS NOT WHETHER NEW YORK SHOULD BUILD AI CAPABILITY.
THE QUESTION IS WHO WILL CONTROL IT — AND IN WHOSE INTEREST.
INDEPENDENT ANALYSIS
This investigation is published independently by Digital Sovereignty Coalition. It does not imply endorsement, sponsorship, institutional affiliation, or partnership with Empire AI or its participating institutions.
Any future formal collaboration must be disclosed and must not compromise DSC's editorial independence.
SOURCING AND EDITORIAL SAFEGUARDS
- Primary institutional documentation is used wherever possible; claims are attributed to the documents that support them.
- Dates, amounts, affiliations, and governance claims are verified against primary sources before being stated as fact. Claims that could not be verified are labeled rather than stated.
- Established facts are separated from analysis, and unanswered questions are identified as questions.
- No quotations, citations, or documents are fabricated. Nothing on this page implies access to confidential information.
- Unverified items are flagged for editorial review and will be updated or removed as verification proceeds.