Provably Secure AI for Health Data – Powered by Confidential Computing

AI is revolutionizing healthcare by optimizing processes and significantly increasing efficiency. However, health data is subject to strict data protection requirements and must not be processed without ensuring patient privacy. The solution lies in confidential computing: this technology makes it possible to keep data fully encrypted, even during processing. With Privatemode AI, for the first time, an LLM solution is available in which prompts are verifiably end-to-end encrypted, without any access by infrastructure, cloud, or service providers. Privatemode makes it possible to securely use state-of-the-art open-source AI models with sensitive data. The architecture ensures that neither the model nor the underlying infrastructure can access the sensitive data. This enables the secure use of LLMs with health data in the cloud, for example in the context of electronic patient records. Confidential computing is already established in the healthcare sector as part of the electronic patient record system "ePA" and now opens up completely new possibilities for reconciling sensitive data with scalable AI applications.