Ransomware is a cybercrime idea as old as the computer itself. The idea is simple: break into a victim’s computer, encrypt all their files, and demand payment for decryption. That being said, the social dynamics at play during this interaction are anything but simple. Generally speaking, the cybercriminal has one goal: to get paid. In order for the victim to pay, they must trust that upon payment they will get their files back. This simple dichotomy gives rise to the Ransomware Trust Paradox

The Paradox

The Ransomware Trust Paradox, an idea coined by Max Smeets in April 2025, refers to the idea that inherently untrustworthy individuals or groups engaging in illegal access to often sensitive data must convince their victims of their trustworthiness to meet their ends (Smeets, 2025). This shows a marked difference between ransomware and other forms of both digital and conventional crime. Think about a bank robbery: both the robber and the ransomware operator have the same goal, to make money, but the bank robber is never trusted by the victim. Rather, they achieve their ends by force. What makes this idea so fascinating to me is the lengths ransomware groups will go to to secure trust from victims. To get an idea for this, let’s first look at the first ransomware, and see what is done differently today.

The First Ransomware

The story of the first ransomware is an interesting one, to say the least. It was created by Joseph Popp, an evolutionary biologist from Harvard working on AIDS research in the 80s. The malware was distributed by mailing floppy discs to attendees to a WHO conference on, believe it or not, AIDS research (O’Kane et al., 2018). Popp was quickly found and arrested, and while no one knows why he did this, his idea led to the creation of an entirely new criminal enterprise in ransomware. In premise, Popp’s ransomware worked on the same basic principle of ransomware today: infect a computer, encrypt everything, and demand payment for it back, but there were several key differences between Popp’s ransomware and modern ransomware. For one, his ransom note (the message telling victims what is happening and what they should do about it) was primitive and nowhere near visually appealing. On top of that, payment was asked to be sent to a PO Box in Panama (O’Kane et al., 2018). There was no clear indication of who the money was being sent to, or how sending cash to Panama would allow one to get their files back. Put yourself in the shoes of a researcher in the 80s. You probably wouldn’t pay in this situation, right? 

Ransomware Today

These days, there are plenty of ransomware groups following in Popp’s footsteps taking various approaches to gaining trust. One ransomware group, Darkside, states plainly in their ransom note that they are here to make money, and if they did not return on their promise to decrypt their victim’s data, no one would pay them going forward (Smeets, 2025). This approach is backed by psychology. Studies have shown that, when marketing to a customer, companies that state their intentions plainly are more likely to gain trust (Dayal, 1999). This extends to the ransomware domain because ransomware operators are also trying to gain trust to receive payment. Another group, BlackCat, engages in anonymous “press releases” to clear up rumors about themselves. In this instance, various news outlets had misattributed attacks to BlackCat, and their “press release” attempted to clarify that they were not making any purely malicious attacks, their sole purpose was to make a deal and return the encrypted data (Smeets, 2025). Another vector of trust groups have exploited rests in User Interface (UI) design. Studies show that just changing the UI in a payment portal increased the number of users who shared their contact information by 31.3% (Yashimi, 2020). This demonstrated that interfaces based on familiar items or other interfaces makes people more likely to give up information (Yashimi et al., 2020). During Conti’s time operating as a ransomware group, they exploited this by using a professional looking ransom note, and sleek payment portals that made victims feel like they were completing a legitimate business transaction. There are many other strategies too, including providing 24/7 customer support, and referring to victims as “customers” (Smeets, 2025).

How Much Damage Does Ransomware Cause?

While it is hard to know exactly how much ransomware is costing year over year, many estimates exist. One estimate from the FBI placed ransomware at a market worth of 200 million USD, with one ransomware family (Cryptowall) totalling 18 million USD in losses in the US alone (O’Kane et al., 2018). How much ransomware operators are asking for in one attack varies by the operator and sometimes by the estimated worth of data encrypted. One company, Symantec, put average ransoms around 679 USD, with some as large as 6285 USD (O’Kane et al., 2018). This of course varies with the price of bitcoin, which is the most popular way for operators to demand payment (Huang et al., 2017). But who is paying these large ransoms? Turns out that, according to a study by Huang et al., 74.5% of Cerber (another family of ransomware) targets were residential. Because these numbers were based on internet service providers, this means that most targets used residential routers, so likely households. While the study is limited in scope due to these numbers being only related to one family of ransomware, these numbers are striking. While only somewhere between 1% and 3% of victims pay the ransom (O’Kane et al., 2018), one can easily see how lucrative the ransomware business is. 

Overall, the Ransomware Trust Paradox is a theory that I find fascinating, the idea that criminal enterprises engaging in explicitly untrustworthy activities are exploiting trust to meet their ends is something never seen in the criminal world before the introduction of ransomware. Tracing ransomware from its roots reveals the importance of this paradox in the operations of modern ransomware groups, and helps us think about how to better defend against them. Going forward, individuals and corporations can use this idea to make better value judgments when they are hit by ransomware, and better understand how to defend themselves against it. 

References

Dayal, S., Landesberg, H., & Zeisser, M. (1999). How to build trust online. Marketing Management, 8(3), 64. Retrieved from https://www.proquest.com/scholarly-journals/how-build-trust-online/docview/194194113/se-2

Huang, D. Y., Aliapoulios, M. M., Li, V. G., Invernizzi, L., Bursztein, E., McRoberts, K., Levin, J., Levchenko, K., Snoeren, A. C., & McCoy, D. (2018, May 1). Tracking Ransomware End-to-end. IEEE Xplore. https://doi.org/10.1109/SP.2018.00047

O’Kane, P., Sezer, S., & Carlin, D. (2018). Evolution of ransomware. IET Networks, 7(5), 321–327. https://doi.org/10.1049/iet-net.2017.0207

Shutterstock. (2024). Ransomware. In Ransomware explained: How it works and how to remove it.

Smeets, M. (2025). The Ransomware Trust Paradox. https://virtual-routes.org/wp-content/uploads/2025/04/Virtual-Routes-Pharos-Report-Series-No.-2.pdf

Yashmi, N., Momenzadeh, E., Taghipour Anvari, S., Adibzade, P., Moosavipoor, M., Sarikhani, M., & Feridouni, K. (2020). THE EFFECT OF INTERFACE ON USER TRUST; USER BEHAVIOR IN E-COMMERCE PRODUCTS. Proceedings of the Design Society: DESIGN Conference, 1, 1589–1596. https://doi.org/10.1017/dsd.2020.103

Odds are, as a consumer you’ve likely glanced at a food label in your lifetime. Though, for some, more than a quick glimpse is necessary, as the intricate details can be the difference between health or illness. This really applies to any allergy, as ingredients and wording can easily veil allergens. Specifically, this can be a regular practice for those avoiding gluten, as just a microscopic amount can have macroscopic repercussions. 

What is gluten?

Gluten is a protein found in wheat, primarily present as gliadin and glutenin. Synonymous proteins are present in rye as secalin, barley as hordein, and oats as avenins (Biesiekierski, 2017). Gluten is used as an umbrella term to describe each of the proteins, as each behave and interact in similar ways. Most commonly found in breads, pastas, and other dough based products, gluten acts as a binding agent. This means its presence impacts texture, elasticity, and even taste (Biesiekierski, 2017). These factors cause gluten to be found in less obvious products like seasonings, dressing, and even the remoistenable seal of an envelope. 

To be or to not be free of gluten

With the possibility of gluten being present in so many different forms, tracking its presence is crucial for those trying to avoid it. As declared by the Food and Drug Administration, for something to be labeled gluten free it must meet specific criteria to be considered not misbranded. The official ruling defines ‘gluten free’ as an item that doesn’t have a gluten containing grain, is not derived from a gluten containing grain, and an item with less than 20 parts per million of gluten (ppm) (Food and Drug Administration, 2013). The parts per million decree is pertinent when a gluten containing grain has been specifically adapted to contain not gluten. It essentially means that for every 1,000,000 grams of a substance, less than 20 of them are able to actually contain gluten. This microscopic amount pales in comparison to a single bread crumb or speck of flour, causing issue to another form of gluten possession, cross contamination. 

Items naturally free of gluten such as meat or rice cannot confidently be deemed gluten free due to cross contamination. Cross contamination occurs when one substance is unknowingly transferred to another, in this case that substance is gluten. Transfer of gluten from a containing substance to supposedly gluten free product can happen at any point in the development process of a product (Damasceno et al., 2024). A single speck of wheat flour would be enough to contaminate a gluten free product. This means shared equipment, such as a conveyor belt holding glutenous food being used on a gluten free product could easily result in the surpassing of the 20 ppm maximum.

External Certifications

Beyond just an FDA approved gluten free marking on a product, there are many different types of certifications that may each follow their own thresholds for what is considered gluten free. Globally, these 3rd party certification processes vary in their guidelines to be included, with different organizations such as Beyond Celiac’s ‘Gluten-Free Certification Program’, Gluten Intolerance Group’s ‘Certified Gluten Free’ Stamp, or National Celiac Association’s gluten free certification. Each of these listings require different ppm of gluten, with one even having a maximum of 5 ppm, which is significantly lower than the FDA regulation (Dennis, 2018). The certifications can provide more comfort to consumers avoiding gluten, as they require more extensive testing of the products to confirm their gluten ppm. While the FDA decree sets limits for a company to legally consider its product gluten free, it does not necessarily require the manufacturer to test the product to confirm that (Thompson & Simpson, 2014). 

How to Read a Label – A Guide from First Hand Experience

As someone with Celiac Disease, with its only treatment being a strict gluten free diet, over the past 9 years of my diagnosis I’ve learned a few tricks along the way. Most notably, I would say being a conscious consumer has been the most impactful thing for my safety when selecting food items. If a product does not have a specific certification, meaning it is  just generally labeled gluten free or has no label at all, my first step is to read the ingredients. In my experience, there are a few specific keywords to look out for. In the process of scanning an ingredient list, if I notice words like wheat, barley, malt, or rye, off the bat the product can be considered not gluten free. Additionally, gluten can be hidden in certain ingredients like corn syrups, oats, yeasts, and some seasonings. These items aren’t inherently glutenous, but particular processes may make them unsafe. 

In some cases, a search online is beneficial for me as well. Frequently, even though a label does not guarantee an item to be gluten free, the company will disclose on their website that their products are free of gluten. Oftentimes, the company will not officially disclose anything due to concern for cross contamination. The actual extent to which this is possible though varies by company. This is where the list at the bottom of a product saying ‘may contain wheat’ is notable. For some, ‘may contain’ could mean direct use of shared equipment, but for others it’s a disclaimer that the factory also makes gluten items. Research or contact with the company is sometimes the best option as it clarifies this uncertainty, and allows for a decision by the consumer depending on the extent of the allergy/sensitivity. Lastly, an item cannot just be deemed gluten free if it lacks a note saying it ‘contains wheat’, as that does not ensure it does not contain barley or rye, which are also gluten containing grains. 

References

Biesiekierski, J. R. (2017, February 28). What is gluten? – biesiekierski – 2017 – Journal of Gastroenterology … Wiley Online Library. http://onlinelibrary.wiley.com/doi/10.1111/jgh.13703/full

Damasceno, R. P. B., Zandonadi, R. P., Mendes, M., Cunha Junior, L. C., Raposo, A., Teixeira-Lemos, E., Chaves, C., & Farage, P. (2024, April 18). Risk of gluten cross-contamination due to food handling practices: A mini-review. MDPI. https://www.mdpi.com/2072-6643/16/8/1198

Dennis, M. (2018). Third Party certification of Gluten Free Foods. BIDMC of Boston. https://www.bidmc.org/centers-and-departments/digestive-disease-center/services-and-programs/celiac-center/celiacnow/nutrition-and-the-gluten-free-diet/label-reading/third-party-certification-of-gluten-free-foods

Food and Drug Administration. (2013, August 5). Food Labeling; Gluten-Free Labeling of Foods. Federal Register – National Archives. https://www.federalregister.gov/documents/2013/08/05/2013-18813/food-labeling-gluten-free-labeling-of-foods

Thompson, T., & Simpson, S. (2014, October 1). A comparison of gluten levels in labeled gluten-free and certified gluten-free foods sold in the United States. Nature News. https://www.nature.com/articles/ejcn2014211 

When a slow, devastating illness collides with the speed and chaos of disaster, the consequences can be dire. Cancer demands constancy—structured protocols, timed treatments, uninterrupted care. Disasters, on the other hand, dismantle. They disrupt and displace, often without warning. As floods rise and conflicts erupt, what becomes of the cancer patient whose chemotherapy is scheduled for tomorrow? Or the child awaiting biopsy results? These questions are no longer hypothetical—they are urgent. We live in a world in which climate-induced events, pandemics, and political instability are increasingly common, and yet cancer care systems remain woefully unprepared (Meara et al., 2015). More alarmingly, disaster preparedness plans globally tend to focus heavily on trauma and communicable diseases, leaving chronic illnesses like cancer out of the picture (Spiegel et al., 2018). This gap in planning and policy is not just an oversight—it is a failure that costs lives. Cancer does not wait. It does not decelerate in times of emergency, and thus our response systems must be re-engineered to reflect that urgency.

Throughout history, the story is tragically familiar. In West Africa during the 2014–2016 Ebola outbreak, already-strained healthcare systems collapsed under the weight of contagion. Oncology units shuttered. Diagnostics were suspended. It is estimated that cancer diagnoses dropped by more than half—not because the disease disappeared, but because access to care did (Lal et al., 2020). During the COVID-19 pandemic, the narrative repeated. Nearly 1.5 million cancer surgeries were postponed or canceled globally in the first phase alone, creating a backlog that continues to impact patients today (COVIDSurg Collaborative, 2020). The disruptions extended beyond surgical delays: patients missed chemotherapy cycles, radiotherapy sessions were halted, and research trials were suspended. Even long after the initial waves subsided, patients continued to face delays due to workforce shortages, overwhelmed hospitals, and financial strain. In low-income countries, the effects were particularly acute—many patients never returned to care at all (Hanna et al., 2020). For diseases where timing can determine survival, such delays are nothing short of devastating.

Not all disasters make global headlines. In Haiti, the slow erosion of healthcare capacity due to political instability has silently decimated oncology care. The country has only one radiotherapy machine, often out of service due to power outages or mechanical failure (IAEA, 2021). Few outside the region know that patients are sometimes forced to travel to the Dominican Republic—an international journey that requires money, paperwork, and time most do not have. In Ukraine, following the 2022 invasion, radiation oncology centers were damaged by shelling, not as deliberate attacks on health services, but as collateral in broader infrastructure strikes. Linear accelerators were damaged, and hospitals had to triage power between radiotherapy equipment and ventilators. These cascading failures don’t just interrupt care—they erase it.

Why is cancer care so fragile in the face of crisis? Unlike many medical disciplines, oncology depends on a convoluted web of components: technology, personnel, pharmaceuticals, and precise timing. Sophisticated machines like PET, MRI, and CT scanners are fixed and highly location-bound. Radiation therapy requires specialized shielding and uninterrupted electricity. Chemotherapy drugs often demand cold-chain logistics and exact calibration. When disasters strike—whether they be cyclones or armed conflict—these systems do not adapt; they fracture (Beddoe, 2019). Health workers are reassigned to trauma units. Oncology wings are converted into ICUs. Supply chains buckle under logistical bottlenecks. Medications expire in warm storage rooms. The loss of a single link can compromise the entire therapeutic chain.

Additionally, health information systems often crash, wiping out records, imaging histories, and continuity plans. Paper records, still common in many parts of the world, are destroyed by flooding or lost in evacuation. Digital systems, where they exist, may lack redundancy or cloud backup (Chan et al., 2022). In such fragile architectures, resilience isn’t just a buzzword—it is a literal matter of life and death.

The price for this failure is rarely paid by the wealthy or politically protected. Rather, the toll is borne by those already standing at society’s margins. It is the refugee with no address, no insurance, and no voting rights. The woman in a remote village with cervical cancer whose screening was never completed. The child born into the shadow of war, whose swollen lymph nodes go ignored until too late. In Syria, Sudan, Gaza, and beyond, cancer becomes the silent casualty of humanitarian disaster (El Saghir et al., 2018). These patients are not just neglected; they are erased from the narrative.

In Lebanon, where millions of Syrian refugees reside, a cancer diagnosis for a non-citizen can be a de facto death sentence due to high costs and lack of access to specialized care (Saab et al., 2019). In Yemen, where humanitarian funding has been slashed, cancer patients are reportedly turning to unproven herbal remedies in desperation. These lives are marginalized not only by disaster but by the systems built to respond to disaster. The harm is not accidental—it is structural.

Less discussed are the internal migrants—those displaced not across borders, but within their own country. After floods in Pakistan displaced over 30 million people in 2022, many oncology patients were unable to access urban hospitals.

Temporary shelters set up for the displaced did not include oncology services. In the United States, Hurricane Maria’s destruction of Puerto Rico’s infrastructure led to cancer patients being airlifted to mainland hospitals—but only those with Medicaid portability or private insurance could access care. Disaster response often assumes temporary needs. But cancer care is not temporary. It cannot be paused and resumed without consequence.

Can we truly plan for the unpredictable? Can we outpace disaster with foresight? The answer, increasingly, is yes. The key lies in data-driven prediction. Artificial intelligence and climate modeling can forecast hotspots of disaster exposure, layered with population health maps to anticipate where cancer care will likely be disrupted. Rwanda, for example, has implemented a national telemedicine network that connects rural health centers with oncology specialists, ensuring continuity of consultation even when physical access is compromised. These are not theoretical innovations—they are operational strategies that can be scaled globally. Predictive algorithms can also identify patients most at risk of losing care continuity, enabling pre-disaster outreach and medication stockpiling. This is anticipatory medicine—a shift from treating the aftermath to engineering the future.

Decentralization is another pillar of resilience. Cancer care must move beyond capital cities and elite hospitals. In India, the expansion of mobile cancer screening and treatment units has proven effective for reaching underserved populations (Pramesh et al., 2014). In Kenya, Project ECHO connects rural clinicians with top oncology experts in real time, democratizing expertise and shrinking the urban-rural divide (Strother et al., 2021). In the Philippines, emergency response plans now incorporate considerations for cancer patients, including medication supply reserves. These models not only deliver care—they redefine what care infrastructure looks like.

Technology can serve as a lifeboat when all else fails. In Rwanda, Zipline drones ferry medical supplies including cancer medications across impassable terrain. In Bangladesh, solar-powered healthcare facilities ensure that diagnostics and treatments can continue even after grid failures during cyclones. Electronic health records with cloud backups ensure that a displaced patient’s oncology file can travel with them, digitally intact, across regions (Guleria & Dua, 2021). These innovations aren’t luxuries; they are future standards. Health systems must invest not just in gadgets but in technology ecosystems that are climate-resilient, secure, and scalable.

For all this progress, the most urgent questions are moral. Who is accountable when patients fall through the cracks? When pharmaceutical embargoes block cancer treatments in war-torn nations, who intervenes? When governments prioritize military spending over oncology units, who is held responsible? (UNDRR, 2015). These are not just questions of policy—they are tests of our ethical architecture. Global health law must evolve to classify cancer care as a protected service during crisis, much like food and shelter. This includes safeguarding cancer infrastructure under international humanitarian law and establishing international corridors for drug delivery in conflict zones (Lancet Commission on Global Access to Pain Control and Palliative Care, 2018; Singh et al., 2020).

The WHO’s Global Initiative for Childhood Cancer shows what coordinated international action can achieve. But such initiatives must expand in both scope and binding authority. Cancer must be built into every emergency preparedness and response framework. Countries must be incentivized—perhaps even mandated—through funding structures and international law to protect cancer services during disaster. Civil society must demand this from their leaders. Humanitarian response teams must include oncologists and palliative care specialists. Cancer cannot be an afterthought.

This is not just about survival—it is about justice. Continuity of cancer care in the face of disaster is a barometer of our civilization. It is a reflection of how we value life, dignity, and equity. The crises ahead—climate displacement, antimicrobial resistance, economic collapse—will test us even more deeply. If we are not ready, we will repeat the same pattern of delay, death, and denial. But if we choose to be ready—if we design for disruption, plan for pain, and invest in resilient care—we can ensure that no diagnosis becomes a death sentence simply because the ground shook, the skies opened, or the sirens wailed. The next disaster is coming. The question is not if—but whether we will meet it with courage or complacency.

References

Beddoe, A. M. (2019). Elimination of cervical cancer: Challenges for developing countries. E-cancer Medical Science, 13, 975. https://doi.org/10.3332/ecancer.2019.975

Chan, E. Y. Y., Huang, Z., Lo, E. S. K., Kim, J. H., Hung, K. K. C., & Shaw, R. (2022). Digital health in the management of non-communicable diseases during and after disasters: A systematic review. International Journal of Environmental Research and Public Health, 19(5), 2777. https://www.mdpi.com/1660-4601/19/5/2777

COVIDSurg Collaborative. (2020). Elective surgery cancellations due to the COVID-19 pandemic: Global predictive modelling to inform surgical recovery plans. British Journal of Surgery, 107(11), 1440-1449. https://doi.org/10.1002/bjs.11746

El Saghir, N. S., Soto Pérez de Celis, E., Fares, J. E., & Sullivan, R. (2018). Cancer care for refugees and displaced populations: Middle East conflicts and global natural disasters. American Society of Clinical Oncology Educational Book, 38, 433-440. https://ascopubs.org/doi/10.1200/EDBK_201365

Guleria, R., & Dua, S. (2021). e-Health in India: Challenges and opportunities. Journal of Family Medicine and Primary Care, 10(1), 151-154. https://www.jfmpc.com/article.asp?issn=2249-4863;year=2021;volume=10;issue=1;spage=151;epage=154;aulast=Guleria

Hanna, T. P., King, W. D., Thibodeau, S., Jalink, M., Paulin, G. A., Harvey-Jones, E., O’Sullivan, D. E., Booth, C. M., Sullivan, R., & Aggarwal, A. (2020). Mortality due to cancer treatment delay: Systematic review and meta-analysis. BMJ, 371, m4087. https://www.bmj.com/content/371/bmj.m4087

International Atomic Energy Agency (IAEA). (2021). Directory of radiotherapy centres. IAEA. https://www.iaea.org/resources/databases/dirac

Lal, A., Erondu, N. A., Heymann, D. L., Gitahi, G., & Yates, R. (2020). Fragmented health systems in COVID-19: Rectifying the misalignment between global health security and universal health coverage. The Lancet, 397(10268), 61-67. https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(20)32228-5/fulltext

Lancet Commission on Global Access to Pain Control and Palliative Care. (2018). Alleviating the access abyss in palliative care and pain relief—an imperative of universal health coverage: The Lancet Commission report. The Lancet, 391(10128), 1391-1454. https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(17)32513-8/fulltext

Meara, J. G., Leather, A. J., Hagander, L., Alkire, B. C., Alonso, N., Ameh, E. A., Bickler, S. W., Conteh, L., Dare, A. J., Davies, J., Mérisier, E. D., El-Halabi, S., Farmer, P. E., Gawande, A., Gillies, R., Greenberg, S. L., Grimes, C. E., Gruen, R. L., Ismail, E. A., … & Yip, W. (2015). Global Surgery 2030: Evidence and solutions for achieving health, welfare, and economic development. The Lancet, 386(9993), 569-624. https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(15)60160-X/fulltext

Pramesh, C. S., Badwe, R. A., Borthakur, B. B., Chandra, M., Raj, E. H., Kannan, T., Kalwar, A., Kapoor, S., Malhotra, H., Nayak, S., Rath, G. K., Sagar, T. G., Sebastian, P., Sarin, R., Shanta, V., Sharma, S. C., Shukla, S., Vijayakumar, M., Vijaykumar, D. K., … & Sullivan, R. (2014). Delivery of affordable and equitable cancer care in India. The Lancet Oncology, 15(6), e223-e233. https://www.thelancet.com/journals/lanonc/article/PIIS1470-2045(14)70117-2/fulltext

Saab, R., Jeha, S., Khalifeh, H., Zahreddine, L., Bayram, L., Merabi, Z., Abboud, M. R., Muwakkit, S., Tarek, N., Haidar, C., Fahd, M., El Solh, H., Al-Ahmari, A., Al-Sweedan, S., Al Sudairy, R., Jastaniah, W., Amayiri, N., Sultan, I., Ghandour, K., … & Al-Rawas, A. H. (2019). Displaced children with cancer in Lebanon: A sustained response to an unprecedented crisis. Cancer, 124(7), 1464-1472. https://acsjournals.onlinelibrary.wiley.com/doi/full/10.1002/cncr.31273

Singh, T., Sharma, S., & Nagesh, S. (2020). Socio-economic impact of COVID-19 pandemic in India. International Journal of Pervasive Computing and Communications, 16(4), 309-319. https://www.emerald.com/insight/content/doi/10.1108/IJPCC-06-2020-0053/full/html

Spiegel, P., Khalifa, A., & Mateen, F. J. (2018). Cancer in refugees in Jordan and Syria between 2009 and 2012: Challenges and the way forward in humanitarian emergencies. The Lancet Oncology, 15(7), e290-e297. https://www.thelancet.com/journals/lanonc/article/PIIS1470-2045(14)70067-1/fulltext

Strother, R. M., Asirwa, F. C., Busakhala, N. W., Njiru, E., Orang’o, E., Njuguna, F., Skiles, J., Carter, J., Mega, A., & Kaspers, G. J. (2021). AMPATH-Oncology: A model for comprehensive cancer care in sub-Saharan Africa. Journal of Cancer Policy, 19, 100282. https://www.sciencedirect.com/science/article/pii/S2213538321000424

United Nations Office for Disaster Risk Reduction (UNDRR). (2015). Sendai framework for disaster risk reduction 2015-2030. UNDRR. https://www.undrr.org/publication/sendai-framework-disaster-risk-reduction-2015-2030

Artificial intelligence (AI) is no longer a futuristic concept confined to research labs—it is embedded in the present and is reshaping the medical world in profound ways. From interpreting X-rays and CT scans with pinpoint precision to forecasting disease risk and suggesting treatment options, AI has quickly established itself as a powerful force in modern healthcare. Its speed and accuracy often surpass those of human doctors, offering clear benefits in diagnostics, efficiency, and cost. However, these advancements come with ethical dilemmas that demand close examination. As machines begin to assume roles once reserved for trained physicians, we must ask: Can—and should—algorithms replace doctors? And if we let them, how do we ensure fairness, accountability, and humanity in the delivery of care?

The appeal of medical AI is undeniable. In 2019, a deep learning model developed by Google Health demonstrated an ability to detect lung cancer in CT scans more accurately than six board-certified radiologists. The AI achieved this by analyzing hundreds of thousands of imaging data points, learning to recognize subtle patterns that human eyes might miss (Ardila et al., 2019). Tools like this are being introduced in hospitals around the world, where they assist in tasks ranging from reading mammograms to diagnosing diabetic retinopathy. According to a report by McKinsey & Company, AI could generate up to $100 billion annually in savings across the U.S. healthcare system by streamlining administrative work, accelerating clinical decision-making, and improving patient outcomes (McKinsey, 2020). Given rising healthcare costs and physician burnout, these gains are attractive to both hospitals and policymakers.

But despite the promising performance of AI in medicine, significant ethical challenges remain. The first, and arguably most urgent, is bias. AI systems are only as good as the data they are trained on, and that data often reflects the deep inequities already present in healthcare systems. A 2019 study published in Science revealed that an algorithm used by millions of patients across the United States underestimated the health needs of Black patients by nearly half. This was because the algorithm used past healthcare spending as a proxy for illness severity, a flawed metric that ignored the systemic barriers many patients of color face when accessing care (Obermeyer et al., 2019). As a result, Black patients received less care—not because they were healthier, but because the system failed to see them clearly. This case is not isolated; other studies have found similar patterns of underdiagnosis and undertreatment when AI models are trained on non-representative datasets.

Another pressing ethical concern is accountability. In traditional clinical settings, doctors are held responsible for their decisions. They must explain their rationale, justify treatment plans, and, when necessary, answer to patients, institutions, and courts. But with AI, the decision-making process becomes opaque. Most advanced machine learning systems—particularly deep neural networks—operate as ‘black boxes,’ making predictions without offering clear explanations. When these systems err, it’s unclear who should be held accountable: the developer who built the model, the healthcare institution that implemented it, or the physician who relied on its recommendations. According to a 2021 report by the U.S. Food and Drug Administration (FDA), regulatory frameworks are still in development, particularly for AI systems that continuously learn and update after deployment (FDA, 2021). Without clear legal standards, patients who are harmed by AI-driven errors may find it difficult to seek justice or even understand what went wrong.

The lack of transparency in AI decision-making also raises concerns about informed consent. Informed consent is a cornerstone of ethical medical practice—it ensures that patients understand and agree to their care plans. But when decisions are made by algorithms that neither patients nor clinicians fully understand, how can consent be truly informed? A study published in The Lancet Digital Health argues that as AI systems become more involved in care, healthcare providers must develop new protocols to explain not just medical options, but the role and limitations of AI in the process (Morley et al., 2020). If patients are unaware that an AI is influencing their diagnosis or treatment—or if they do not understand the implications—they may be making decisions without full understanding, violating their autonomy.

Beyond bias and accountability, there is a deeper concern: the potential loss of human connection in healthcare. Medicine is not just about science; it is also about relationships, empathy, and trust. A 2022 survey by the AMA found that over 80% of patients believe that a strong relationship with their doctor is critical to their health outcomes (AMA, 2022). Doctors not only interpret data—they read body language, offer emotional support, and provide continuity of care. These are aspects of medicine that AI, no matter how sophisticated, cannot replicate. An algorithm can analyze a heart rate trend, but it cannot notice the anxiety in a patient’s voice or offer comfort during a terminal diagnosis. If AI becomes the default decision-maker, we risk reducing patients to data points and care to computation.

However, the goal of integrating AI into medicine is not necessarily to replace doctors—but to augment them. When thoughtfully designed and ethically implemented, AI can be an invaluable assistant. It can take on time-consuming tasks like sorting lab results, triaging patient records, or flagging high-risk cases, allowing clinicians to spend more time with their patients. A study published in Nature Medicine supports this collaborative model, showing that physicians using AI-assisted tools made more accurate diagnostic decisions than either AI or humans alone (Topol, 2019). This suggests that the future of medicine doesn’t belong to AI or doctors—it belongs to both. The question is how to ensure that this collaboration works in a way that protects ethical standards and centers the patient’s well-being.

To achieve this balance, several changes are needed. First, developers must ensure that AI is trained on diverse, representative data. This includes not only demographic diversity but also data reflecting various healthcare systems, cultural backgrounds, and comorbid conditions. Second, transparency and explainability must be built into AI design. Algorithms used in healthcare should be able to provide understandable reasoning behind their recommendations so that doctors and patients can question and interpret their output. Third, regulatory bodies must develop robust legal frameworks for AI in medicine, including rules for post-deployment monitoring, auditability, and liability assignment. Finally, medical education must evolve. Doctors need to be trained not only to use AI tools, but to critically evaluate them—understanding when to trust the machine and when to override it.

In conclusion, the integration of AI into medicine offers enormous potential. It promises faster diagnoses, reduced errors, and broader access to high-quality care. But it also introduces serious ethical challenges that cannot be ignored. Bias, accountability, transparency, and the erosion of human connection are not side issues—they are central to the future of ethical, patient-centered care. As we design and deploy these systems, we must ensure that they serve all patients equitably, support rather than supplant clinicians, and maintain the humanity at the heart of medicine. AI can be a transformative force—but only if we are equally committed to the ethics that guide its use. The future of healthcare is not about choosing between man or machine—it’s about finding the right way for both to work together.

References

American Medical Association. (2022). Patient Trust and the Doctor-Patient Relationship in the Digital Age. https://www.ama-assn.org

Ardila, D., Kiraly, A. P., Bharadwaj, S., et al. (2019). End-to-end lung cancer screening with three-dimensional deep learning on low-dose chest computed tomography. Nature Medicine, 25(6), 954–961. https://doi.org/10.1038/s41591-019-0447-x

McKinsey & Company. (2020). How artificial intelligence will impact healthcare. https://www.mckinsey.com

Morley, J., Machado, C. C. V., Burr, C., Cowls, J., Joshi, I., Taddeo, M., & Floridi, L. (2020). The ethics of AI in health care: A mapping review. Social science & medicine (1982), 260, 113172. https://doi.org/10.1016/j.socscimed.2020.113172

Obermeyer, Z., Powers, B., Vogeli, C., & Mullainathan, S. (2019). Dissecting racial bias in an algorithm used to manage the health of populations. Science, 366(6464), 447–453. https://doi.org/10.1126/science.aax2342

Topol, E. J. (2019). High-performance medicine: the convergence of human and artificial intelligence. Nature Medicine, 25(1), 44–56. https://doi.org/10.1038/s41591-018-0300-7

U.S. Food and Drug Administration. (2021). Artificial Intelligence/Machine Learning (AI/ML)-Based Software as a Medical Device (SaMD) Action Plan. https://www.fda.gov/media/145022/download