OpenMed Agent Achieves End-to-End Local Medical Inference Loop on iPhone with Full Privacy Protection
OpenMed Agent's medical AI Agent implements nearly full local processing on the iPhone, culminating in a call to GPT-5.5.
This Agent has read-only access to Apple Health data (heart rate, blood pressure, sleep, blood oxygen, blood sugar, etc.), supports scanning or photographing medical documents, completes all identifier desensitization locally on the device, and then uses the user's own ChatGPT account to call GPT-5.5 for medical inference, referencing real authoritative sources such as Mayo Clinic and Cleveland Clinic.
Source: Public Information
ABAB AI Insight
Maziyar Panahi has previously focused on local AI and privacy computing. This time, OpenMed Agent constructs a strict privacy loop through device-side actions and user-owned API keys, addressing the core pain point of data leakage in the implementation of medical AI, similar to the edge processing concept of Apple Intelligence.
In terms of capital pathways, OpenAI and medical AI applications are accelerating privacy-first edge deployments, with funding concentrating on personal medical Agent platforms that support local inference, health data integration, and strong compliance capabilities, significantly enhancing ordinary users' medical privacy protection levels.
This design is akin to the early extension of local voice assistants into professional medical scenarios. Currently, personal health AI is undergoing a significant transformation from cloud data dependence to device-side privacy-first inference.
Essentially, this represents a restructuring of the industry chain: medical data processing and inference are migrating from centralized cloud platforms to local user devices, as edge desensitization and zero-identifier leakage mechanisms greatly reduce regulatory risks and user concerns, shifting pricing power from cloud AI reliant on large-scale data uploads to edge solutions supporting secure local inference.
ABAB News · Cognitive Law
Data must stay on the device for medical AI to truly integrate into personal life; privacy is not a limitation but a prerequisite for medical intelligence. Local actions combined with user-owned keys are the correct way to unlock privacy and capability; the smarter the cloud, the less users dare to use it. True personal AI never demands all your data but is responsible for you only within the scope you allow.