By Chris Hoofnagle
Buried in a 1991 campus policy with an admirable title, Policy Guidelines Governing Openness and Freedom to Publish, is a blanket prohibition on an entire category of scholarship. The policy is right about a great deal: It bars extramural sponsor from veto power over what a faculty member publishes, and it correctly treats that guarantee as central to the University’s mission.
But in the middle of a document dedicated to protecting the freedom to publish, the campus forbids its faculty from all funded classified research:
The Berkeley campus does not accept classified research projects from any sponsor, even if such classification is intended only to permit project personnel access to classified information and/or entry into classified research sites. The principal reasons that classified projects are unacceptable are (1) the resultant requirement for a campus facility clearance and (2) the inherent publication restrictions. In general, classified projects are not consistent with the teaching, research, and public service missions of the Berkeley campus.
There are serious arguments for caution about classified work on a university campus, and I take them up below. Yet the three specific rationales the policy offers do not withstand scrutiny, even as written in 1991, and beneath them lies a deeper conceptual error: the policy invokes the freedom to publish as grounds for forbidding research a faculty member has freely chosen to do.
A freedom turned into a prohibition
Not every restriction is an enemy of freedom. Freedom has its paradoxes. The policy’s central rule, that a sponsor may not veto a faculty member’s publications, is a valuable restriction, because it shields the scholar’s autonomy from outside control. Scholars also take on genuine limits in the ordinary course of their work: obligations to human subjects, the terms attached to proprietary datasets, promises of confidentiality. These are compatible with and even constitutive of responsible academic freedom.
The classified-research ban is different in kind. The sponsor-veto rule restricts others’ power over a faculty member’s work; the professional obligations are duties the scholar assumes within work she has chosen. The classified ban, by contrast, is the institution forbidding the choice itself.
Rationale one: the facility clearance
The policy’s first stated reason is administrative: that classified work would require “a campus facility clearance.” A facility clearance is indeed a demanding, externally audited security regime, and it is fair to say that an open, densely trafficked campus is not the natural home for compartmented work. That is, in fact, why the University long ago sited its most sensitive classified research off-campus at the labs.
But that observation supports a narrow, location-based policy: find an appropriate facility. The campus manages secure environments routinely. University Health Services holds protected medical records under federal privacy law and physical safeguards. The law school’s clinics operate as a working law practice, with segmented, access-controlled space for privileged client files. Laboratories across campus handle hazardous biological agents and controlled substances under federal regimes that require background vetting of personnel and controlled, inspected facilities. The campus even operates an animal-research facility whose location and entry are deliberately restricted.
Rationale two: inherent publication restrictions
The second reason is the one closest to the policy’s stated purpose: classified work carries “inherent publication restrictions.” This is true, but an incomplete and inconsistent view.
The policy treats bounded publication limits as categorically disqualifying only here, while accepting them elsewhere in the same document. The guidelines permit sponsors to delay publication for review and they accommodate restrictions on the dissemination of specified proprietary materials. The policy, in other words, already recognizes that “freedom to publish” coexists with real, negotiated limits on timing and content.
Classification does not seal knowledge off. Methods, instincts, and insights developed on controlled work routinely inform a scholar’s open work, and vice versa. Boundaries are porous in a free society.
Classified fields, moreover, sustain their own scholarly literatures. The CIA’s Studies in Intelligence is a peer-reviewed journal running both classified and unclassified articles; the NSA’s Cryptologic Quarterly serves a comparable function for signals intelligence. “Classified” does not mean “outside the community of scholarly criticism.” It means a differently bounded community, one that many distinguished researchers have moved within and beyond throughout their careers, maintaining cutting-edge open programs alongside sensitive government work.
Rationale three: consistency with the University’s missions
The third reason is the broadest and the most deeply mistaken: that “classified projects are not consistent with the teaching, research, and public service missions of the Berkeley campus.” This sweeping claim is the hardest to reconcile with the historical record and common sense.
Consider the technologies that reshaped ordinary life over the past century. Silicon Valley itself is the product of hundreds of billions of advanced research funding through the DOD, giving birth to untold advances in materials sciences, chemistry, and physics. Digital computing traces its origin to wartime codebreaking; it was the Army and Navy that generously funded early computing in order to solve artillery and other problems. Satellite navigation began as a military system before it guided civilian life.
Classified and defense programs have repeatedly seeded the most consequential public benefits, a pattern difficult to square with the assertion that such work is inconsistent with a research or public-service mission. Even your use of this website and e-commerce itself is made possible by academic cryptographers who moonlighted for the intelligence community.
The breadth of classified research is enormous, even unfathomable, because the needs of the different military and intelligence branches span the ocean floor to outer space. Nor is the reach confined to the physical sciences. Classified work also draws on the humanities and social sciences: languages, history, and area studies, because governments must understand other peoples.
The nature of the policy’s claim is especially bizarre given the history of UC and Berkeley. The University of California has administered Los Alamos and Lawrence Livermore. JASON, the elite defense advisory group whose reports are largely classified, counted Berkeley physicists among its founders.
Whether one likes this work or not, it is difficult to maintain that the University regards classified national-security research as categorically inconsistent with its public-service mission when it has treated the administration of exactly such research as a signal public service for decades.
Moving forward
The ban on classified research is not costless. The Department of Defense is the federal government’s largest single funder of research and development. Its annual budget for research, development, test, and evaluation now approaches $145 billion, roughly triple the entire budget of the National Institutes of Health. A substantial share of that work is classified, and Berkeley’s policy forecloses it entirely. Consider Georgia Tech as a peer—it maintains extensive facilities for classified work and in 2024 ranked second in the nation in federally sponsored research expenditures (over $1.4bn compared to our ~800mm). We cannot compete.
Berkeley need not seek out classified research, and there remain real reasons for a public university to approach it with care. Classified work can create a two-tier community of the cleared and the uncleared. It can complicate the training of graduate students. It genuinely constrains publication. A serious policy could weigh all of this and still decide, on balance, to keep most classified work at arm’s length or off-campus.
But those are not the reasons the policy gives. The rationales articulated in 1991 by then–Provost for Research Joseph Cerny were mistaken when written and remain mistaken today: the campus can and does run secure facilities; bounded publication limits are already tolerated across the University’s own research; and the claim that such work betrays the University’s mission is refuted by the University’s own history.
Berkeley’s guidelines get openness right in nearly every other respect. On classified research, they would be stronger for saying plainly what concerns actually motivate the position and for ceasing to justify a limit on faculty choice in the name of the very freedom that choice is meant to express.