Technology for a hospital in exchange for patients’ medical information
Sheba Medical Center has signed a strategic agreement with OpenAI to integrate AI models into its clinical workflows. This partnership highlights the value of Israel's centralized medical data while raising critical questions about patient privacy and data security.
The Sheba Medical Center has signed a strategic cooperation agreement with OpenAI, granting medical teams access to the artificial intelligence giant's specialized medical models. This agreement may prove to be one of the most significant collaborations between Israel’s healthcare system and an international technology leader since the vaccine partnership with Pfizer during the pandemic.
Israel offers a unique asset: a centralized, digital healthcare system rich in medical information. For AI developers, this comprehensive data repository, accumulated over decades, serves as essential raw material for developing and calibrating accurate medical models.
Innovation and Integration
Sheba is the first hospital outside the United States to gain access to OpenAI’s dedicated medical AI tools, primarily ChatGPT Health. By integrating institutional knowledge—such as clinical protocols and treatment pathways—directly into the model, the system can generate recommendations tailored to the Israeli population and the specific working environment of its doctors.
For Sheba’s ARC innovation center, this early access aims to:
-
Streamline bureaucratic processes
-
Reduce the administrative burden of medical documentation
-
Enhance personalized medicine and continuity of care
However, the partnership raises complex ethical and legal questions regarding data ownership. Even when information is anonymized, the use of patient data by commercial entities without explicit consent (Opt-in) risks eroding public trust in the healthcare system.
Cyber Risks and Clinical Safety
The Israeli Ministry of Health has previously identified public AI tools as a significant cyber threat, leading to attempts to restrict their use in government hospitals. Despite these measures, many medical professionals continue to use external AI tools via personal devices, potentially bypassing secure networks and risking the leakage of sensitive patient information.
Beyond privacy, there is the clinical risk of AI "hallucinations," where models present false information as fact. Unsupervised reliance on these tools could lead to diagnostic errors. The primary challenge for regulators is to establish a framework that leverages the benefits of AI while ensuring strict data protection and mitigating cyber risks.
