Sam Cooper

Shared Labs, Shared Harm names MIT, Oxford and McGill among universities working with Beijing-backed AI institutes linked to Uyghur repression and China’s security services.

Over the past five years, some of the world’s most technologically advanced campuses in the United States, Canada and the United Kingdom — including MIT, Oxford and McGill — have relied on taxpayer funding while collaborating with artificial-intelligence labs embedded in Beijing’s security state, including one tied to China’s mass detention of Uyghurs and to the Ministry of Public Security, which has been accused of targeting Chinese dissidents abroad.

That is the core finding of Shared Labs, Shared Harm, a new report from New York–based risk firm Strategy Risks and the Human Rights Foundation. After reviewing tens of thousands of scientific papers and grant records, the authors conclude that Western public funds have repeatedly underwritten joint work between elite universities and two Chinese “state-priority” laboratories whose technologies drive China’s domestic surveillance machinery — an apparatus that, a recent U.S. Congressional threat assessment warns, is increasingly being turned outward against critics in democratic states.

The key Chinese collaborators profiled in the study are closely intertwined with China’s security services. One of the two featured labs is led by a senior scientist from China Electronics Technology Group Corporation (CETC), the sanctioned conglomerate behind the platform used to flag and detain Uyghurs in Xinjiang; the other has hosted “AI + public security” exchanges with the Ministry of Public Security’s Third Research Institute, the bureau responsible for technical surveillance and digital forensics.

The report’s message is blunt: even as governments scramble to stop technology transfer on the hardware side, open academic science has quietly been supplying Chinese security organs with new tools to track bodies, faces and movements at scale.

It lands just as Washington and its allies move to tighten controls on advanced chips and AI exports to China. In the Netherlands’ Nexperia case, the Dutch government invoked a rarely used Cold War–era emergency law this fall to take temporary control of a Chinese-owned chipmaker and block key production from being shifted to China — prompting a furious response from Beijing, and supply shocks that rippled through European automakers.

Cameras and Drones

The Strategy Risks team focuses on two state-backed institutes: Zhejiang Lab, a vast AI and high-performance computing campus founded by the Zhejiang provincial government with Alibaba and Zhejiang University, and the Shanghai Artificial Intelligence Research Institute (SAIRI), now led by a senior CETC scientist. CETC designed the Integrated Joint Operations Platform, or IJOP — the data system that hoovered up phone records, biometric profiles and checkpoint scans to flag “suspicious” people in Xinjiang.

United Nations investigators and several Western governments have concluded that IJOP and related systems supported mass surveillance, detention and forced-labor campaigns against Uyghurs that amount to crimes against humanity.

Against that backdrop, the scale of Western collaboration is striking.

Since 2020, Zhejiang Lab and SAIRI have published more than 11,000 papers; roughly 3,000 of those had foreign co-authors, many from the United States, United Kingdom, and Canada. About 20 universities are identified as core collaborators, including MIT, Stanford, Harvard, Princeton, Carnegie Mellon, Johns Hopkins, UC Berkeley, Oxford, University College London — and Canadian institutions such as McGill University — along with a cluster of leading European technical universities.

Among the major U.S. public funders acknowledged in these joint papers are the National Science Foundation (NSF), the National Institutes of Health (NIH), the Office of Naval Research (ONR), the Defense Advanced Research Projects Agency (DARPA) and the Department of Transportation. For North America, the warning is twofold: U.S. and Canadian universities are far more entangled with China’s security-linked AI labs than most policymakers grasp — and existing “trusted research” frameworks, built around IP theft, are almost blind to the human-rights risk.

In one flagship example, Zhejiang Lab collaborated with MIT on advanced optical phase-shifting — a field central to high-resolution imaging systems used in satellite surveillance, remote sensing and biometric scanning. The paper cited support from a DARPA program, meaning U.S. defense research dollars effectively underwrote joint work with a Chinese lab that partners closely with military universities and the CETC conglomerate behind Xinjiang’s IJOP system.

Carnegie Mellon projects with Zhejiang Lab focused on multi-object tracking and acknowledged funding from the National Science Foundation and the U.S. Office of Naval Research. Multi-object tracking is a backbone technology for modern surveillance — allowing cameras and drones to follow multiple people or vehicles across crowds and city blocks. “In the Chinese context,” the report notes, such capabilities map naturally onto “public security applications such as protest monitoring,” even when the academic papers present them as neutral advances in computer vision.

The report also highlights Zhejiang Lab’s role as an international partner in CAMERA 2.0, a £13-million U.K. initiative on motion capture, gait recognition and “smart cities” anchored at the University of Bath, and its leadership in BioBit, a synthetic-biology and imaging program whose advisory board includes University College London, McGill University, the University of Glasgow and other Western campuses.

Meanwhile, SAIRI has quietly become a hub for AI that blurs public-security, military and commercial lines.

Established in 2018 and run since 2020 by CETC academician Lu Jun — a designer of China’s KJ-2000 airborne early-warning aircraft and a veteran of command-and-control systems — SAIRI specializes in pose estimation, tracking and large-scale imaging.

Under Lu, the institute has deepened ties with firms already sanctioned by Washington for their roles in Xinjiang surveillance. In 2024 it signed cooperation agreements with voice-recognition giant iFlytek and facial-recognition champion SenseTime, as well as CloudWalk and Intellifusion, which market “smart city” policing platforms.

SAIRI also hosted an “AI + public security” exchange with the Ministry of Public Security’s Third Research Institute — the bureau responsible for technical surveillance and digital forensics — and co-developed what Chinese media billed as the country’s first AI-assisted shooting training system. That platform, nominally built for sports, was overseen by a Shanghai government commission that steers AI into defense and public-security applications, raising the prospect of its use in paramilitary or police training.

Outside the lab, MPS officers have been charged in the United States with running online harassment and intimidation schemes targeting Chinese dissidents, and MPS-linked “overseas police service stations” in North America and Europe have been investigated for pressuring exiles and critics to return to China.

Meanwhile, Radio-Canada, drawing on digital records first disclosed to Australian media in 2024 by an alleged Chinese spy, has reported new evidence suggesting that a Chinese dissident who died in a mysterious kayaking accident near Vancouver was being targeted for elimination by MPS officers and agents embedded in a Chinese conglomerate that the U.S. Treasury accuses of running a money-laundering and modern-slavery empire out of Cambodia.

The new reporting focuses on a former undercover agent for Office No. 1 of China’s Ministry of Public Security — the police ministry at the core of so-called “CCP police stations” in global and Canadian cities, and reportedly tasked with hunting dissidents abroad.

Taken together, cases of alleged Chinese “police station” networks operating globally, new U.S. Congressional reports on worldwide threats from the Chinese Communist Party, and the warnings in Shared Labs, Shared Harm suggest that Western universities are not only helping to build China’s domestic repression apparatus with U.S. taxpayer funds, but may also be contributing to global surveillance tools that can be paired with Beijing’s operatives abroad.

To counter this trend, the paper urges a reset in research governance: broaden due diligence to weigh human-rights risk, mandate transparency over all international co-authorships and joint labs, condition partnerships with security-linked institutions on strict safeguards and narrow scopes of work, and strengthen university ethics bodies so they take responsibility for cross-border collaborations.

By Nicolas Hulscher, MPH

The single largest vaccine–dementia study ever conducted (n=13.3 million) finds risk intensifies with more doses, remains elevated for a full decade, and is strongest after flu and pneumococcal shots.

The single largest and most rigorous study ever conducted on vaccines and dementia — spanning 13.3 million UK adults — has uncovered a deeply troubling pattern: those who received common adult vaccines faced a significantly higher risk of both dementia and Alzheimer’s disease.

The risk intensifies with more doses, remains elevated for a full decade, and is strongest after influenza and pneumococcal vaccination. With each layer of statistical adjustment, the signal doesn’t fade — it becomes sharper, more consistent, and increasingly difficult to explain away.

And critically, these associations persisted even after adjusting for an unusually wide range of potential confounders, including age, sex, socioeconomic status, BMI, smoking, alcohol-related disorders, hypertension, atrial fibrillation, heart failure, coronary artery disease, stroke/TIA, peripheral vascular disease, diabetes, chronic kidney and liver disease, depression, epilepsy, Parkinson’s disease, cancer, traumatic brain injury, hypothyroidism, osteoporosis, and dozens of medications ranging from NSAIDs and opioids to statins, antiplatelets, immunosuppressants, and antidepressants.

Even after controlling for this extensive list, the elevated risks remained strong and remarkably stable.

Vaccinated Adults Had a 38% Higher Risk of Dementia

The primary adjusted model showed that adults receiving common adult vaccines (influenza, pneumococcal, shingles, tetanus, diphtheria, pertussis) had a:

38% increased risk of developing dementia (OR 1.38)

This alone dismantles the narrative of “vaccines protect the brain,” but the deeper findings are far worse.

Alzheimer’s Disease Risk Is Even Higher — 50% Increased Risk

Buried in the supplemental tables is a more shocking result: when the authors restricted analyses to Alzheimer’s disease specifically, the association grew even stronger.

50% increased risk of Alzheimer’s (Adjusted OR 1.50)

This indicates the effect is not random. The association intensifies for the most devastating subtype of dementia.

Clear Dose–Response Pattern: More Vaccines = Higher Risk

The authors ran multiple dose–response models, and every one of them shows the same pattern:

Dementia (all types)

From eTable 2:

• 1 vaccine dose → Adjusted OR 1.26 (26% higher risk)
• 2–3 doses → Adjusted OR 1.32 (32% higher risk)
• 4–7 doses → Adjusted OR 1.42 (42% higher risk)
• 8–12 doses → Adjusted OR 1.50 (50% higher risk)
• ≥13 doses → Adjusted OR 1.55 (55% higher risk)

Alzheimer’s Disease (AD) Shows the Same—and Even Stronger—Trend

From eTable 7:

• 1 dose → Adjusted OR 1.32 (32% higher risk)
• 2–3 doses → Adjusted OR 1.41 (41% higher risk)
• ≥4 doses → Adjusted OR 1.61 (61% higher risk)

This is one of the most powerful and unmistakable signals in epidemiology.

Time–Response Curve: Risk Peaks Soon After Vaccination and Remains Elevated for Years

Another signal strongly inconsistent with mere bias: a time-response relationship.

The highest dementia risk occurs 2–4.9 years after vaccination (Adjusted OR 1.56). The risk then slowly attenuates but never returns to baseline, remaining elevated across all time windows.

After 12.5 years, the risk is still meaningfully elevated (Adjusted OR 1.28) — a persistence incompatible with short-term “detection bias” and suggestive of a long-lasting biological impact.

This pattern is what you expect from a biological trigger with long-latency neuroinflammatory or neurodegenerative consequences.

Even After a 10-Year Lag, the Increased Risk Does Not Disappear

When the authors apply a long 10-year lag — meant to eliminate early detection bias — the elevated risk persists:

• Dementia: OR 1.20
• Alzheimer’s: OR 1.26

If this were simply “people who see doctors more often get diagnosed earlier,” the association should disappear under long lag correction.

Influenza and Pneumococcal Vaccines Drive the Signal

Two vaccines show particularly strong associations:

Influenza vaccine

• Dementia: OR 1.39 → 39% higher risk
• Alzheimer’s: OR 1.49 → 49% higher risk

Pneumococcal vaccine

• Dementia: OR 1.12 → 12% higher risk
• Alzheimer’s: OR 1.15 → 15% higher risk

And again, both exhibit dose–response escalation — the hallmark pattern of a genuine exposure–outcome relationship.

Taken together, the findings across primary, supplemental, dose–response, time–response, stratified, and sensitivity analyses paint the same picture:

• A consistent association between cumulative vaccination and increased dementia risk

• A stronger association for Alzheimer’s than for general dementia

• A dose–response effect — more vaccines, higher risk

• A time–response effect — risk peaks after exposure and persists long-term

• Influenza and pneumococcal vaccines strongly drive the signal

• The association remains after 10-year lag correction and active comparator controls

This is what a robust epidemiologic signal looks like.

In the largest single study ever conducted on vaccines and dementia, common adult vaccinations were associated with a 38% higher risk of dementia and a 50% higher risk of Alzheimer’s disease. The risk increases with more doses, persists for a decade, and is strongest for influenza and pneumococcal vaccines.

Nicolas Hulscher, MPH

Epidemiologist and Foundation Administrator, McCullough Foundation

Support our mission: mcculloughfnd.org

Please consider following both the McCullough Foundation and my personal account on X (formerly Twitter) for further content.