Artificial intelligence is already visibly changing healthcare — from diagnostics through telemedicine to the training of future doctors. Alongside the benefits, however, come new risks: from fake videos and ethical dilemmas to cyberattacks on hospitals. What can AI really do today, and what does that mean for us?
When AI helps doctors
One test showed that a language model, without special medical training, was able to answer questions from the US Medical Licensing Examination with a success rate of approximately 50–60 %, while the passing threshold is around 60 %. Importantly, it had to not only choose the correct option, but also explain why it was correct and why the others were incorrect, which is beneficial for students. A 2020 study, in turn, reported that in diagnosing tumors on X-ray images, AI achieved a success rate of roughly 88 %, whereas radiologists scored about 75 %. Added to this are telemedicine and prevention based on conversations and data, which save patients trips and increase the efficiency of facilities.
The dark side and necessary defense
Attackers are using the same techniques that are speeding up medicine. According to statements from January 2024, security teams are already unable to keep up with the rise in AI-related cyber incidents. AI can search for vulnerabilities in environments, recognize behavioral patterns, and even generate code to attack devices. Meanwhile, healthcare is increasingly interconnected, yet suffers from a shortage of people to continuously cover these threats.
Specific to AI are attacks on datasets and model manipulation — even small, unobtrusive interventions can lead to skewed diagnoses or improperly developed drugs. A human alone can hardly track hundreds of threat sources and dozens of models in real time. It is therefore sensible to involve artificial intelligence in defense as well: for monitoring, analysis, and prediction of what is happening and what may happen. The sooner this becomes a routine part of practice, the better.
