They are also at risk from backdoor attacks, data poisoning and unauthorised access. 

Chatbots are now outperforming human healthcare workers in answering patient inquiries, streamlining processes and reducing administrative burdens. 

But hospitals should balance their use of artificial intelligence (AI) with cybersecurity risks given that the sector is a primary target of attacks, according to experts. 

Generative AI (genAI) and human responses to patient questions did not differ statistically in terms of accuracy, completeness, and relevance, according to a study by New York University Grossman School of Medicine released in July. 

GenAI responses outperformed human providers in terms of understandability and tone by 9.5%, were more than twice as likely (125%) to be considered empathetic and 62% more likely to use language that conveyed positivity, it added. 

Fullerton Health Philippines in Bonifacio Global City (BGC) is one of those healthcare providers that have started using a chatbot through its LiveFuller app to ease human staff load. 

“The application can give a patient’s health condition and provide a predictive analysis of diseases likely to be acquired if habits are not changed,” Carmie de Leon, Electronic Health System head at Fullerton BGC, said in a Zoom interview with HealthCare Asia. 

Clients get their diagnostic results faster, she said. “The biggest help is to staff members who don’t need to encode tons of documents, allowing them to better focus on providing quality services.” 

Reynalyn Tomada, director for operations at Fullerton BGC, noted that whilst genAI has gone a long way in healthcare, it could be hindered by cybersecurity risks. 

Temus Pte Ltd., which provides IT services, technology, and consultancy to public and private sectors to fast-track their digital transformation, said electronic health systems that use genAI are at risk from backdoor attacks, data poisoning and unauthorised access. 

“The second privacy risk is the potential leakage of personally identifiable information from electronic health record data into LLMs (large language models) during training,” Temus Managing Director for AI Matt Johnson and Dickon Smart-Gill, data and enterprise architect for strategy, said in a joint emailed reply to questions. 

LLM is a type of machine learning model that can generate and classify text, answer questions conversationally, and translate text from one language to another. 

Geoffrey Coley, regional chief technology officer for Asia-Pacific at Veritas Technologies LLC, noted that as cybercrimes remain unabated, “patients are increasingly concerned about data privacy, including how their medical records are being handled.” 

Healthcare is a key target of cyberattacks, he said, citing IBM’s Cost of Data Breach Report released in July. 

The average breach cost for healthcare fell 10.6%, to $9.77 million in the 12 months through February 2024. “But that factor wasn’t enough to remove it from the top costliest industry for breaches — a spot it’s held since 2011,” IBM said. 

Still, analysts are optimistic about genAI’s potential in healthcare. 

Johnson and Smart-Gill noted that for treatment, genAI could enhance rather than replace clinical judgement, giving doctors data-driven insights for a more informed decision. In research, genAI could transform how people understand health. 

“By analysing large volumes of anonymised health data, researchers could identify patterns in diseases and treatment outcomes. This could lead to the generation of new hypotheses for research and accelerate medical discovery,” they added.