Healthcare just won a Nobel Prize for AI.
Now it has to defend against it too.

In October 2024, the Nobel Prize in Chemistry went to an AI system for the first time. Half went to Demis Hassabis and John Jumper of Google DeepMind for AlphaFold2, which predicts a protein's 3D structure from its amino acid sequence with roughly 90% accuracy — solving a problem biology had chased for fifty years. The other half went to David Baker for computational protein design: using AI to invent protein structures that don't exist in nature, many with direct therapeutic potential. AlphaFold's database is now free to more than two million researchers in 190 countries and covers over 200 million protein structures — essentially every protein science has sequenced. This isn't a healthcare AI pilot. It's a permanent shift in how medicine discovers what's biologically possible.

The same technology, pointed the other way

Deepfake attacks across all sectors rose roughly 880% in 2024, and healthcare has become one of the sharpest edges of that curve. Mount Sinai researchers found that AI-generated fake X-rays were convincing enough to fool practising radiologists — even when the readers knew in advance that some images in the set were synthetic. Attackers are now using cloned video and voice to impersonate patients well enough to secure telehealth prescriptions for controlled substances, and to impersonate physicians or administrators convincingly enough to talk staff into releasing confidential patient records over the phone. In April 2026 the American Medical Association issued a formal policy framework calling for protections against deepfake impersonation of physicians, warning explicitly that synthetic audio and video can mislead patients, influence clinical decisions, and erode trust in care itself.

Where AI is advancing medicine — and where it now has to be governed

This is a workforce problem before it's a technology problem

The global market for AI in precision medicine alone is projected to grow from roughly $4 billion in 2026 to more than $125 billion by 2040. That growth curve only translates into better outcomes if the clinicians using these tools are trained on two fronts at once: how to act on an AI-generated prediction, and how to recognise when a clinical image, voice or identity in front of them has been fabricated. Those aren't separate skills anymore — they're both part of clinical judgment now, the same way sterile technique became part of surgical judgment. I led nearly 3,000 clinicians through Abu Dhabi's Covid-19 response and reached over 78,000 through DoH Abu Dhabi's ICU e-learning programme, including a 24-module course built specifically for communicable-disease response. Training at that scale is exactly the mechanism that closes an AI-literacy gap before it becomes an AI-fraud incident — and it's the same discipline behind AI Governance: know what the system can prove, not just what it predicts.