Cybersecurity vulnerabilities in healthcare systems are increasing as cybercriminals target institutions that typically operate with fragmented security standards and outdated software. Brendan S. Kelly, an expert in AI and pediatric radiology, highlighted a significant cyberattack in March 2021 on Ireland's National Health Service (HSE), where ransomware spread through a phishing email, causing wide-scale disruptions and patient safety concerns. Experts at a recent conference discussed the additional vulnerabilities introduced by AI-powered large language models (LLMs), which blur the lines between instructions and data, making it easier for non-experts to execute cyberattacks. Techniques such as "prompt injection" pose serious threats to diagnostic imaging data. Professor Renato Cuocolo emphasized that damage from data poisoning is challenging to remediate as it requires fully retraining AI models. He also warned about the risks of generative AI producing synthetic data too closely resembling real patient information, creating a new potential access point for attackers. To mitigate these risks, experts advocate for cybersecurity measures such as the principle of least privilege, sandboxing AI models, and implementing digital watermarking for data verification. Additionally, there is a call for increased human oversight and training to better prepare healthcare staff for emerging cyber threats.
Thu, 02 Apr 2026 04:03:12 GMT | healthcare-in-europe.com