Request a demo

Mapping the Research Security Landscape

Interactive research security landscape

Mapping the Research Security Landscape

30 universities. 92,301 publications. 1,581 institutional relationships.

The Department of War’s audit order raised an immediate question: what does the collaboration landscape actually look like?

This interactive analysis maps three decades of publications involving the 30 universities and entities identified through the Section 1286 research-security framework — revealing where relationships are concentrated, how the research mix has shifted, and which relationships matter most.

Explore the findings
Skip to the data explorer

1995–2026 publication record · 2026 partial

Open sources only · unclassified

265 entities on the Section 1286 list

47 Shared or transferred patents

The map shows

Observable relationships in public publication, patent, and researcher-affiliation records.

The map supports

Prioritization, exploration, and contextual review.

The map does not determine

Whether an individual relationship was improper, restricted, or inconsistent with a requirement.

Finding 01

A vast network - with a clear center of gravity.

1,581 university–entity relationships appear in the record — yet just 57 whole institutions from the 1286 list appear at all — plus 43 named labs inside them — and a handful of relationships drive most of the volume.

0

Distinct institutional relationships in the record

0

institutions + 43 labs

Whole institutions from the 1286 list in the record, plus named labs inside them

0

Publications in the highest-volume observed relationship

0

Entity-linked publication records for the largest 1286 entity

The heaviest relationships are the natural starting point — and the small, strategically significant ones stand out on their own.

Where the volume sits

Switch views and select any row to open its detail.

Author count
All publications included. Roughly 9.9% carry more than 50 authors.

    Counts overlap and must not be summed.

    A single publication can involve several institutions and several 1286 entities at once, so it is counted in each. Institution and entity attributions run through 2025 and sum to 124,103 and 125,207 respectively, against 90,482 distinct publications in that window; the 92,301 headline adds a partial 2026. Volume describes how much material sits behind a relationship — it is not a score and not a ranking of institutions. Two coverage facts matter here: only 55.3% of author records carry any affiliation, so every count is a floor; and about 9.9% of publications have more than 50 authors, which inflates some figures considerably. Use the author-count control above to see the difference.
    Text alternative & full data table
    Concentration in the publication record
    Finding 02

    The collaboration landscape is not simply growing. Its technical composition is changing.

    Between 2015 and 2025 the published mix shifted decisively toward critical technologies — AI up 325%, space up 415%.

    Where the Research Landscape Is Moving

    Strategically relevant publication tags, 2010–2026. Select topics to compare how the published research mix has changed. 2026 is in progress, so its counts are partial.

    Topics Shift-click any topic to view it alone — even one greyed out at the five-topic limit.

    Topic tags overlap, and a topic label does not establish that a publication contained controlled information or served a defense purpose.

    These are analytical tags applied to publications, not a partition of them: one publication can carry several tags and many carry none, so the counts cannot be added into a publication total. Select any topic name below the chart to read a plain-language explanation of what the area includes, why it matters, and what a tag does not establish. Indexed mode sets each topic's first year with recorded activity to 100.
    Text alternative & full annual data table

    Read a topic in plain language

    Every one of the 14 areas has a short explanation of what it includes, why it matters, and what the data does not establish.

    Finding 03

    The public record reveals more than co-authorship.

    Patents, researcher affiliation patterns, and institutional name resolution each add a different kind of evidence to the picture.

    Joint patent ownership is a more discrete and documentable relationship than co-authorship — negotiated terms and an ownership claim exist on the public record. It is not evidence of impropriety.

    ORCID covers approximately 5% of author-ID rows, so researcher-affiliation figures should be treated as a lower bound.

    Accurate entity resolution is essential because academic names and acronyms frequently collide — two institutions in this group share the acronym USC, and three separate institutions worldwide are called Northeastern University.

    Every layer of this map rests on entity resolution — matching 265 entities and their name variants across languages and transliteration systems. That discipline is the foundation of Exiger’s risk platform, and it is what makes a review defensible.

    Layer 02

    Explore the Landscape

    Six views over the same record. Select anything to open its detail.

    UNIVERSITIES

    The 30 universities in scope

    Select any institution for its names, relationships, and record. The Department's release did not enumerate the institutions; this roster comes from cited DefenseScoop reporting.

      ENTITIES RELATIONSHIPS TOPICS PATENTS RESEARCHERS
      This view was built entirely from the public record. Exiger extends it with an institution’s own agreements, funding, and disclosure data.
      Request a briefing
      Read the methods first
      Methods and sources

      What sits behind the map

      Enough to judge the analysis on its merits. Every figure shown was recomputed from the source workbook.

      Three kinds of relationship, all drawn from the public record:

      • Publication collaboration. A publication carrying at least one author affiliated to a university in scope and at least one affiliated to a 1286 entity.
      • Patent relationships. A patent on which a university in scope and a 1286 entity both appear as applicant or owner.
      • Researcher-affiliation patterns. One researcher with a persistent author identifier appearing on both sides, sequentially or in the same year.

      The dataset does not cover funding sources, grants and subawards, gifts, partnership agreements, joint laboratories, licensing, data-sharing or visiting-scholar arrangements, export-control determinations, or whether any publication led to a transfer of technology. Those answers live in institutional records.

      • Co-authorship is not evidence of funding, technology transfer, misconduct, or an export-control violation.
      • Counts by institution, entity, country, relationship, field, and topic overlap and must not be summed.
      • 92,301 is the deduplicated publication headline (including a partial 2026); every other total double-counts by design.
      • Topic tags overlap, and a tag does not establish that a publication contained controlled information or served a defense purpose.
      • The workbook supports institution × topic and entity × topic views, but not relationship × topic. No individual relationship is linked to a topic anywhere here.
      • Researcher-affiliation figures rest on ORCID identifiers and are a lower bound.
      • Only 55.3% of author records carry any affiliation, so all counts are floors rather than measurements.
      • About 9.9% of collaboration publications have more than 50 authors. Large physics and genomics consortia inflate some figures substantially, so key views offer an excluding large consortia comparison.
      • Repeated titles cap duplicate-record inflation at roughly 4.5%.

      Tier A means a whole-institution identifier match — the entity is an institution in its own right. Tier B means a laboratory or subunit matched on affiliation text, which requires manual validation before anything rests on it. Generic Tier C names were excluded.

      The analytical crosswalk contains 265 institutions, laboratories, subunits, and name variants; the current FY2025 Section 1286 list contains 130 institutions. These universes have not yet been fully reconciled, so this analysis should not be treated as a formal compliance determination.

      The 30 universities come from DefenseScoop’s reporting of 17 August 2026; the Department’s own release did not name them. All 30 reconcile exactly with the institutions in the workbook.

      • Department of War, release, 17 August 2026The audit order: scope, required actions, entity basis, and the 31 August 2026 reporting date.
      • DefenseScoop, 17 August 2026The roster of 30 notified universities, which the Department’s release did not itself publish.
      • Department of War, release, 23 July 2026The FY2025 Section 1286 list of 130 academic and research institutions in China, Russia, and Iran.
      • Analytical workbookEvery quantitative figure, computed from scholarly publication and patent records sourced via Lens.org.
      • University name-reference tableOfficial names, acronyms, alternate and former names, and name-collision cautions for the 30 institutions.
      • Section 1286, FY2019 National Defense Authorization ActThe statutory basis for the Department’s list of foreign institutions of concern.
      Tagged