Last week, the Health Ministry decided to proactively block access to external AI tools, including ChatGPT, Gemini and Claude, from computers connected to government hospital networks.
The move followed a directive the ministry issued in March amid heightened cyber threats facing Israel's healthcare system during Operation Rising Lion.
Gallery
The Health Ministry decided to proactively block access to external AI tools
(Photo: shutterstock)
The ministry has now tightened those restrictions by actively blocking access to the tools. While the measure may reduce the Health Ministry's liability as a regulator for negligent use, it does not actually prevent physicians from using AI. They can continue accessing the tools on their personal smartphones or laptops.
"I carry my laptop everywhere because so many websites are blocked in hospitals," a physician at one of Israel's hospitals told ynet. "Today I can't even access my personal email from the hospital, so I've gotten used to bringing my own computer. I also use chatbots and all those programs, some of which I pay for, through my laptop."
He said he does not use AI tools for clinical diagnosis but believes they offer significant advantages for medicine and research.
"Medicine is heading in that direction. These tools improve output and dramatically reduce the time spent on tasks," he said. "Today it takes me a third of the time to prepare a presentation and a tenth of the time to write a research grant application. I run everything through AI."
'The block solves nothing'
The Health Ministry said it is leading efforts "to integrate dedicated security solutions that will enable broader use of artificial intelligence, including on computers connected to organizational networks, while maintaining the highest standards of privacy, information security and cyber protection."
Prof. Shahar Shelly, a faculty member at the Technion – Israel Institute of Technology Faculty of Medicine, head of its Laboratory for Artificial Intelligence and Decision Making and director of the Neurology Department at Rambam Health Care Campus, said using external free AI tools involves two distinct risks: cyber threats and clinical risks. The current restrictions, he said, fully address neither.
Using external free AI tools involves two distinct risks: cyber threats and clinical risks
(Photo: shutterstock)
"The first is a direct attack vector, where a hostile actor exploits the organization's network to penetrate its systems. The block does prevent that," he said. "The second is the leakage of sensitive information, and here the block doesn't help. A physician who wants to consult such a tool will simply upload patient information another way, such as from a phone, and the data will leak through an unmonitored channel."
He added that the restrictions also fail to address clinical risk.
"A physician relying on AI could receive incorrect medical advice. Blocking access through the hospital network doesn't change that. Instead, it pushes the most sensitive clinical use outside the system's field of vision, which is a disadvantage."
"The impact on physicians who want to use these tools is close to zero," Shelly said. "A network-level block doesn't manage risk. It merely shifts it. The system removes responsibility from itself and places it on physicians, without protecting against the underlying risk."
Shelly said most current use of large language models is focused on work surrounding medicine rather than bedside clinical decision-making.
"They're used to draft discharge letters, translate documents, summarize articles, prepare responses for medical committees and sometimes assist with diagnosis. At that level, the block is frustrating and reduces physicians' effectiveness."
AI use among physicians is rising
Artificial intelligence has become increasingly common in healthcare in recent years, driven by staffing shortages and growing workloads.
Prof. Shahar ShellyPhoto: Rambam Medical CenterA survey published on the American Medical Association website in February 2025 found that physicians' use of AI for specific tasks nearly doubled in just one year, while enthusiasm for the technology continues to grow despite lingering concerns.
According to the survey, nearly two-thirds of responding physicians (66%) reported using AI in healthcare in 2024, up from 38% in 2023. Physicians reported using AI for visit documentation, discharge instructions, treatment plans, translation and diagnostic support.
Some 35% said their enthusiasm for AI outweighed their concerns, compared with 30% a year earlier. Meanwhile, the share of physicians whose concerns exceeded their enthusiasm fell from 29% in 2023 to 25% in 2024. Two out of five physicians said they felt equal measures of enthusiasm and concern.
Several Israeli hospitals already operate internal AI tools integrated into their organizational networks to assist physicians. Shelly said those systems are useful for relatively simple tasks but fall short when it comes to more complex challenges.
"An internal tool can draft an email, translate or summarize an article. But what we need to overcome the challenges of modern medicine, increasing rates of chronic illness, longer life expectancy and ever more complex care, are tools that support decision-making."
According to Shelly, the Health Ministry should not be asking whether physicians should be allowed to use AI tools. Instead, it should define the clinical standard for using them.
"The ministry needs to determine what constitutes proper versus improper use and specify what information may or may not be shared. You can't distinguish between careful and negligent use until you've defined what proper practice looks like. Blocking access is the default response of someone who doesn't know how to manage the risk, so they simply cut off access."
He argued that AI represents a deeper shift in medicine.
"For centuries, medicine was based on the idea that clinical judgment resided with one person, the physician. AI is no longer just another source of information. It participates in the cognitive process itself. It can formulate a diagnosis and organize the reasoning behind it. The block is a reflexive response from a system that senses this change but lacks the tools to control it."
Shelly stressed that the technology will not be stopped by the current restrictions.
The ministry's decision also sparked debate on social media
(Photo: shutterstock)
"It can't be stopped. People already use AI for everything, from choosing a restaurant meal to planning a trip, yet medicine is falling behind," he said. "The only way to deal with this is to allow its use under measurable oversight: define what is permitted, how compliance is measured and what the appropriate standard is. Governance based on prohibition is weaker than governance based on standards."
He said cooperation between the Health Ministry and major technology companies is needed to establish standards for the use of AI in medicine.
'My WhatsApp Is full of patient information'
The ministry's decision also sparked debate on social media.
"Three in the morning. The resident is exhausted. A complex patient arrives and they immediately photograph the medical record on their personal phone. No cybersecurity, no privacy, no quality control. Is this really what you intended by allowing AI use only from personal devices?" wrote Prof. Ido Wolf, director of the Oncology Division at Tel Aviv Sourasky Medical Center (Ichilov Hospital) and head of the School of Medicine at Tel Aviv University, on X.
One commenter wrote: "This is a joke. My WhatsApp is full of CT videos containing patient information. All the trauma groups are full of labels with patient details. If anything, it would have been better to secure these AI tools inside the hospital with what's called an enterprise chatbot. But that requires planning and money, so it's easier just to block them."
Another commenter questioned the policy, noting that "all of the Health Ministry's professional development workshops over the past two years have focused on how to use AI tools to improve our work and performance."
Even ChatGPT has an opinion
So what does ChatGPT think of the new restrictions? It also argued that AI use should be regulated rather than banned.
"In my view, a blanket ban on physicians' use of artificial intelligence would be an extreme step, though restrictions on certain uses are justified," it wrote.
Alongside AI's clear advantages, including saving time, reducing administrative burdens and assisting with diagnosis, ChatGPT noted real risks. These include the possibility that AI systems may make mistakes or "hallucinate" information, physicians may become overly reliant on AI recommendations instead of exercising independent medical judgment and unregulated use could compromise patient privacy.