The Download: supercharged scams and studying AI healthcare

The Double-Edged Sword of AI: How Generative Intelligence Is Reshaping Scams, Healthcare, and Global Power

In late 2022, the world got its first real taste of generative AI’s potential—and its peril. When ChatGPT burst onto the scene, it didn’t just impress with its ability to write poetry or debug code; it revealed a new frontier where machines could mimic human thought with startling accuracy. But as quickly as innovators celebrated this leap forward, cybercriminals saw an opportunity. What began as a tool for creativity and productivity has now become a weapon in the hands of bad actors, fueling a surge in AI-driven scams that are faster, smarter, and harder to detect than ever before.

At the same time, artificial intelligence is quietly transforming healthcare—offering doctors digital assistants, automating diagnoses, and even predicting patient risks before symptoms appear. Yet beneath the promise lies a troubling question: Are these tools actually making patients healthier? Or are we automating care without understanding the consequences?

This is the paradox of our AI moment: a technology capable of healing and harming in equal measure, reshaping industries, societies, and the very fabric of trust.

The Rise of AI-Powered Cybercrime: When Machines Learn to Deceive

The evolution of cybercrime has always mirrored technological progress. From the early days of phishing emails riddled with spelling errors to today’s hyperrealistic voice clones and AI-generated fake websites, attackers have continuously adapted. But the advent of large language models (LLMs) like ChatGPT, Claude, and now DeepSeek-V4 has supercharged their capabilities.

Gone are the days when scammers relied on poorly written emails from “Nigerian princes.” Today, AI can generate flawless, context-aware phishing messages tailored to individual victims—down to their job title, recent purchases, or even emotional state. These aren’t just generic spam blasts; they’re personalized lures crafted in seconds, at scale.

Cybercriminals are also using AI to automate vulnerability scanning—probing corporate networks for weaknesses at speeds no human team could match. Some are even deploying deepfake videos in corporate espionage, creating fake board meetings or executive announcements to manipulate stock prices or extract sensitive data.

The economics are irresistible. Running an AI scam campaign is now cheaper and faster than ever. A single hacker can launch thousands of tailored attacks with minimal effort, while traditional defenses—like spam filters and antivirus software—struggle to keep up with the sheer volume and sophistication.

Deepfakes and the Erosion of Trust

Perhaps the most insidious use of AI in cybercrime is the rise of deepfakes—AI-generated audio, video, or images that are nearly indistinguishable from reality. These aren’t just tools for political disinformation (though they’ve been used for that too); they’re increasingly being weaponized in personal scams.

Imagine receiving a video call from your boss asking you to urgently wire funds to a new vendor. The voice, the mannerisms, even the background—all match perfectly. Only later do you discover it was a deepfake. These “vishing” (voice phishing) attacks are becoming alarmingly common, especially in industries like finance and healthcare where trust is paramount.

The implications go beyond financial loss. Deepfakes threaten to erode trust in all digital communication. If you can’t believe your eyes or ears, how do you verify reality? This “liar’s dividend”—the ability to dismiss real evidence as fake—could undermine everything from journalism to courtroom testimony.

Healthcare AI: Promise vs. Proof

While cybercriminals exploit AI’s mimicry skills, the medical field is betting on its analytical power. AI tools are now embedded in hospitals worldwide, assisting with everything from clinical documentation to radiology.

For example, AI-powered scribes listen to doctor-patient conversations and automatically generate structured medical notes, reducing administrative burden. Other systems scan electronic health records (EHRs) to flag patients at risk of sepsis, heart failure, or diabetic complications—sometimes days before symptoms appear.

Radiology is another frontier. AI algorithms can detect tumors in X-rays and MRIs with accuracy rivaling—or even exceeding—that of human radiologists. A 2022 study in The Lancet found that an AI system reduced missed lung cancer cases by 20% compared to traditional screening.

But here’s the catch: accuracy doesn’t equal impact.

Why? Because healthcare isn’t just about detecting disease—it’s about treatment, follow-up, and human connection. An AI might flag a high-risk patient, but if the clinic is understaffed or the patient lacks transportation, that insight may never lead to action. Moreover, many AI tools are trained on biased datasets, leading to disparities in care for marginalized communities.

The Accountability Gap in Medical AI

One of the biggest challenges in healthcare AI is transparency. Most diagnostic algorithms are “black boxes”—even their developers don’t fully understand how they arrive at certain conclusions. This lack of explainability makes it difficult for doctors to trust or challenge AI recommendations.

Regulators are playing catch-up. The FDA has approved over 700 AI-based medical devices since 2015, but critics argue that many were cleared through fast-track processes without rigorous real-world testing. Meanwhile, liability remains unclear: If an AI misdiagnoses a patient, who’s responsible—the doctor, the hospital, or the software company?

Today’s systems face similar hurdles. Without clear evidence that AI improves survival rates, reduces hospitalizations, or enhances patient satisfaction, widespread adoption may stall—even if the technology itself is sound.

Global Tensions and the AI Arms Race

As AI reshapes industries, it’s also fueling geopolitical rivalry. The U.S. and China are locked in a high-stakes competition for AI dominance, with national security, economic power, and technological sovereignty on the line.

Recent developments highlight the tension. DeepSeek, a Chinese AI firm, just launched DeepSeek-V4—an open-source model that reportedly rivals the capabilities of OpenAI’s GPT-4 and Google’s Gemini. What makes it remarkable is its optimization for Huawei chips, sidestepping U.S. export restrictions on advanced semiconductors.

Meanwhile, the White House has accused Chinese firms of “mass AI theft,” alleging that they’re reverse-engineering American models to accelerate their own development. Beijing denies the claims, calling them “slander” and part of a broader effort to stifle China’s technological rise.

This isn’t just about who builds the best chatbot. AI underpins everything from military drones to financial markets. Control over AI means influence over the future of global power.

The Education Dilemma: Should AI Be in Schools?

As governments grapple with AI’s role in warfare and healthcare, a quieter debate is unfolding in classrooms. Should students use AI tools for learning—or are they just digital crutches?

Some educators worry that AI-powered tutors and essay generators encourage cheating and undermine critical thinking. In response, several countries are restricting access: Norway has banned social media for children under 15, and the Philippines is considering similar measures. In the U.S., a growing movement is pushing to remove AI from schools altogether, arguing that it distracts from foundational skills.

But others see AI as a democratizing force—a way to provide personalized tutoring to students in under-resourced schools. Tools like Khanmigo (an AI tutor from Khan Academy) adapt to each student’s pace and learning style, offering explanations, quizzes, and feedback in real time.

The challenge is balance. Banning AI outright may leave students unprepared for a tech-driven workforce. But unchecked use risks creating a generation dependent on machines for basic reasoning.

The Path Forward: Ethics, Regulation, and Human Judgment

So where do we go from here? The answer lies not in rejecting AI, but in governing it wisely.

We need stronger regulations to prevent AI misuse—like mandatory watermarking of synthetic media, stricter data privacy laws, and international agreements on AI warfare. At the same time, we must invest in “AI literacy” so that doctors, teachers, and citizens can understand and question these tools.

In healthcare, randomized controlled trials should become the gold standard for evaluating AI tools—just as they are for new drugs. We need to measure not just accuracy, but real-world impact: Do patients live longer? Feel better? Receive more equitable care?

And in cybersecurity, we must shift from reactive defense to proactive resilience. That means training employees to spot AI scams, deploying AI-powered detection systems, and fostering a culture of skepticism—where every unexpected request is double-checked.

Ultimately, the future of AI isn’t predetermined. It will be shaped by the choices we make today—about transparency, equity, and the kind of world we want to build. The technology is here. The question is whether we’ll let it serve us—or deceive us.

This article was curated from The Download: supercharged scams and studying AI healthcare via MIT Technology Review