When CEOs and scientists wax poetic about the potential benefits of AI, the technology’s impact on human health often rises to the top—and in many ways, they’re not wrong. Artificial intelligence’s ability to sift through massive datasets makes these algorithms ideal for spotting early-stage cancer or discovering novel drugs. The United Nations even notes that AI-powered disease surveillance could detect potential outbreaks quickly and prevent crises from spiraling into global pandemics.
However, like any tool, artificial intelligence can also be used for nefarious purposes, and a new perspective published in the New England Journal of Medicine highlights a particularly insidious one and its potential impacts on human health: mass surveillance. Debates over how governments and companies collect our personal data are usually framed as civil liberties or national security issues. But experts from Massachusetts General Hospital, Johns Hopkins University School of Medicine, and Harvard Medical School say that being watched can make us sick, and in the age of AI, things could get much, much worse.
The authors identify at least three ways surveillance harms health. First, knowing (or even believing) that you’re constantly being monitored can lead to hypervigilance and chronic “anticipatory stress,” which takes a toll on both physical and mental health. Second, surveillance can scare vulnerable people away from essential health behaviors. This includes groups with a reasonable fear of persecution, such as undocumented immigrants or people seeking reproductive or gender-affirming care, who may avoid pharmacies and clinics, hold back details from their doctors, or even skip buying groceries. Third, surveillance data can be used to hand out things that shape our health, like insurance, housing, and jobs, unequally.
“Historically, mass surveillance programs in the United States have most often been organized by government agencies and subject to democratic oversight,” the authors write. “AI-enabled surveillance, however, relies on data ecosystems that are controlled primarily by private companies.”
That’s a big distinction. The Fourth Amendment’s protections are limited when people voluntarily share information with third parties, even if they don’t realize they’re doing so. HIPAA also generally doesn’t cover data collected by consumer-facing tech companies. As a result, AI companies and other private actors can access and analyze vast amounts of personal data, and they could share it with the government without your knowledge or permission. Traditional surveillance usually needed human oversight. AI, by contrast, can continuously comb through social media, search results, financial transactions, geolocation records, and camera footage, then combine it all into increasingly powerful predictions.
This concern isn’t without real-world precedent. The Denver-based Palantir Technologies, founded by billionaire Peter Thiel, developed AI-enabled tools for U.S. Immigration and Customs Enforcement (ICE), including a system known as ImmigrationOS. ICE has used these tools to combine information from multiple government databases, including visa histories and Medicaid records, to target people for immigration enforcement. Another tool, called Enhanced Leads Identification and Targeting for Enforcement (ELITE), pulls names, photos, addresses, and race and ethnicity data from Medicaid records to map areas with people who could be detained, according to reporting in The BMJ. It can reportedly even generate a “confidence score” for whether someone might be at home.
For the authors, all this surveillance amounts to public health problem. When data collected for health care or social services can be repurposed for surveillance, trust erodes, and people may stop using vital programs. That’s why they argue that AI-enabled mass surveillance should be considered an emerging “structural determinant of health,” which they define as a social, political, environmental, or economic condition that shapes how health and illness are distributed across a population.
“We believe medicine has an urgent responsibility to confront these developments,” the authors write. “Decisions about AI-enabled mass surveillance are being made with limited input from health experts, even though these systems have clear consequences for population health.” Health care technologies can contribute to the problem, too. Clinicians increasingly recommend wearables and smartphone apps that continuously collect patient data, which means the medical system may be feeding the same data streams that could be used for surveillance.
Still, there are ways to fight back. The authors recommend that health care institutions push policymakers for stronger health data protections. They point to Washington state’s 2023 My Health My Data Act, which extended privacy protections to some consumer health data that HIPAA doesn’t cover, including information inferred by algorithms. Nevada and Connecticut have passed similar laws. But the authors warn that these protections are limited, differ from state to state, and don’t stop government agencies from getting and using that data.
The courts offer another option. More than 20 state attorneys general have sued the federal government for sharing Medicaid data with ICE and Palantir, and the authors say health care institutions can support these lawsuits by publicly opposing troubling data-sharing deals. Clinicians can also help by documenting patients’ fears about how their data is used and any resulting changes in whether they seek care, which gives litigators and policymakers evidence to work with.
AI surveillance could still be a boon to public health. The same ability to rapidly pull together data from different sources that makes it dangerous could also help health officials spot and respond to threats more effectively. In the end, AI is just a tool, and it’s up to us how we use it. But the authors warn that we may not have long to decide. By the time AI surveillance’s effects become fully apparent, its infrastructure “may already be deeply embedded in everyday life.” As they put it, “The norms established today could determine whether AI-enabled surveillance becomes a tool in service to population health or a driver of avoidable health inequities and structural harm.”
Darren lives in Portland, has a cat, and writes/edits about sci-fi and how our world works. You can find his previous stuff at Gizmodo and Paste if you look hard enough.