The Patient Control Imperative
Big Tech wants your medical records. Don’t hand over the keys.
CVS just joined OpenAI, Anthropic, Amazon, Microsoft, and Perplexity in announcing health-labeled versions of large language model AI. A recent NY Times article summarized the situation and discussed the benefits and risks. These mainstream medical AI chatbots are doing the same job based on the same health records as our Medical AI Assistant, MAIA, presented in my previous post.
So why would anyone bother with an open-source, private medical chatbot?
The answer lies in who controls your data, who controls your doctor, and who ultimately controls your care.
The Illusion of Privacy and the Data Grab
Consider that six of the largest consumer AI businesses have already announced health-labeled AI. These for-profit, proprietary businesses need every patient to share all of their records with them. AI desperately needs authentic data from reputable sources for training. While these companies can legally absorb your data under HIPAA without consent by claiming it is “de-identified,” but de-identified data cannot be combined across multiple sources like insurance, pharma, and hospitals.
To get the full picture—including highly sensitive details about your life and social circumstances—they need your explicit consent.
This creates a dangerous dynamic. Hospitals and Electronic Health Record (EHR) vendors have aligned interests in using AI to control both physicians and patients. Conversely, if physicians and patients are to maintain control of our own AI, the data has to bypass the institutional EHR and come directly from the patient.
Reclaiming Your “Digital Twin”
The literature and media now talk about creating a “digital twin”—a virtual model of a physical object. While digital twins were originally used to reduce maintenance costs for machinery like locomotives, patients are not machines. We patients must insist that we choose, pay for, and completely control our digital twin.
Imagine sitting in an exam room. Your doctor is forced to type into an employer-mandated AI that optimizes for the hospital’s billing codes and throughput. Meanwhile, your comprehensive, private medical history sits unused. This paradigm shift to patient control is only practical if doctors accept your patient-controlled AI as authoritative, rather than defaulting to the AI forced upon them by their employers.
The 21st Century Cures Act supports this autonomy. It prohibits the FDA from regulating medical decision support and mandates that patients have computational access to their own records. This means you don’t have to trust closed, corporate AI models with your private records.
The Changing Economics of Medical Software
Unlike physical goods, software and knowledge are “non-rivalrous.” If I give you a piece of software, it doesn’t mean I have less of it to use myself. However, corporate AI threatens to create a massive imbalance by locking this knowledge behind paywalls.
To ensure a fair and open ecosystem:
Generative AI must be open source, allowing anyone to inspect and improve it.
Data centers must be substitutable, preventing any single tech giant from holding a monopoly.
Institutions must allow choice, letting individuals use the AI tools they bring to work.
We are already seeing the economics of software development change. AI-assisted coding (sometimes called “vibe-coding”) is making open-source software effectively free to produce. Because creating software is becoming so accessible, licensed domain experts like physicians no longer have to be locked into buying rigid, proprietary systems from EHR vendors. They can actively shape and contribute to the open-source tools they use to treat patients.
MAIA: A Methodology, Not a Device
Our open-source project, MAIA, demonstrates how patients and physicians can control their own AI. Crucially, MAIA is not a medical device; it is a published methodology.
By defining it this way, we shift the power from corporate device manufacturers back to the medical community:
A Secure Filing System: Instead of a complex “vector database” it uses, think of MAIA as a private, highly secure filing system that instantly connects the dots across your entire medical history.
Open Weights: The underlying AI model is completely open, meaning independent experts can verify exactly how it reasons and makes connections.
Peer Review over FDA Clearance: When a physician adopts this methodology, they are practicing medicine informed by published, peer-reviewed literature, rather than operating a black-box medical device. This provides a level of safety and independent verification that standard device classification cannot match.
The Endgame: Holding the Keys to Your Data
The final, largely ignored missing piece of an open internet is how we give people the digital keys to control their own data. The corporate “Agents” currently pushed by Big Tech are simply an attempt to front-run people’s realization that they need to own their own authorization systems.
Ultimately, you need to be the one holding the keys. You need a system capable of setting the rules for who gets to see your data, when they see it, and what they can do with it. Rather than relying on corporate terms of service, user communities—whether patient advocacy groups or medical sub-specialties—will self-assemble to create and share privacy policies that fit their specific needs.
Patients must not give up the choice of private AI that they pay for and control. Relying on open-source software, guided by communities relevant to your individual context, is the only way to preserve autonomy.
Take Control Today
If you want to see what it looks like to be the project manager of your own biology, you can explore the open-source MAIA repository on GitHub or request a temporary, hosted account to see the methodology in action with your own records.