The clinical AI company UpDoc recently announced that the U.S. Food and Drug Administration (FDA) had cleared its AI platform as software as a medical device using patient-facing large language models for proactive, personalized care delivery and coordination.  

The company stated that its platform is the first FDA-cleared agentic clinical AI platform designed to support, rather than replace, doctors, and that the system enables clinicians to deploy FDA-cleared AI agents integrated within a health system’s electronic health records.  

The company and its CEO, Sharif Vakili, MD, MBA, MS, have been careful to describe the technology as “clinical AI,” or AI that delivers care previously requiring a licensed clinician, rather than an “AI doctor.”  

“Clinical AI does not replace the doctor or pretend to be one. It instead augments the doctor’s care and allows patients to receive ongoing continuous care, instead of just waiting for the next appointment. This has never before been possible without human intervention,” Dr. Vakili said in a post on LinkedIn. “Crossing that threshold changes everything. Safety becomes paramount. Auditability and accountability are non-negotiable.”  

"We fundamentally believe that clinical AI should be held to the highest standard," Dr. Vakili said in the press release. "This FDA clearance marks a historic milestone for the field and the beginning of a new era in clinical care, where clinicians will be able to deploy clinical AI agents to help complete tasks autonomously on their behalf." 

“This breakthrough also has the potential to make primary care viable again by directly addressing some of the health-care system's most pressing challenges: patient access, rising system costs, and clinician burnout,” Dr. Vakili added.  

UpDoc’s platform functions as a clinical operating system, enabling AI agents to coordinate with patients and deliver care between clinic visits. Capabilities include adjusting medications, modifying doses, facilitating lab orders, and more.  

"Chronic disease management has always been constrained by the episodic nature of care delivery. The next frontier is making care continuous with smarter coordination between visits. Clinical AI that embeds meaningfully into how care teams work has the potential to bend the curve on some of the most persistent challenges in population health," said Desi Kotis, PharmD, chief pharmacy executive at UCSF Health and vice dean of clinical affairs at the UCSF School of Pharmacy. 

The platform will initially be deployed at Cleveland Clinic, Allegheny Health Network, and UCSF Health.  

Digging Deeper 

UpDoc received FDA 510(k) premarket clearance in December 2025 for a mobile application and web portal designed to help manage patients with type 2 diabetes by optimizing their insulin intake. The cleared device includes both patient and health-care provider interfaces and is classified as a drug dosing calculator.  

The company obtained clearance by demonstrating that its technology was substantially equivalent to a marketed predicate device: the d-Nav Insulin Guidance System, which received FDA clearance in 2019.  

In the FDA documents, the company states that the devices/applications are similar in their software-based insulin dose calculator, software components, ability to connect to blood glucose meters, and other core functions. However, UpDoc notes that its technology differs in its voice and chat interfaces. Patients can enter data through voice or chat, including by interacting with an UpDoc agent that collects information about glucose values, treatment adherence, and symptoms. The software can also guide patients to seek medical attention or follow other safety-related instructions, as appropriate.  

The company also specified that the system “does not interpret or diagnose symptoms” but enables patients to log symptoms that are considered when generating treatment plan instructions. The FDA submission also notes that no clinical testing was performed; instead, the company conducted software, cybersecurity, and human factor testing.  

Based on the FDA submission, the large language model does not appear to drive clinical decision-making but instead functions as an interface for collecting patient data. Health tech reporters have noted that the CEO has declined to disclose additional details about how the large language model operates within the platform, citing competitive reasons.  

Additionally, the clearance includes a Predetermined Change Control Plan that allows certain post-clearance modifications without requiring additional FDA review. These changes include revising default settings and clinical definitions that do not alter the clinical decision logic, expanding existing dosing workflow options, enhancing the user interface, incorporating alternative data input methods, and adding support for additional insulin formulations.  

Ongoing Questions 

Even with greater clarity about what the FDA cleared, broader questions remain about how regulators should evaluate clinical and autonomous AI.  

Many commentators have examined the regulation of generative AI (such as Blumenthal et al), particularly as regulatory agencies approach the threshold of clearing AI systems that function as clinical decision-makers. Is FDA clearance of medical devices sufficient for adaptive, intelligent systems that continue to evolve over time, or is a new regulatory paradigm needed? If so, should responsibility remain with the FDA, or should another body oversee the approval of generative AI in a clinical setting? Some have argued that these platforms should be licensed similarly to clinicians (eg, Bergman et al, Bressman et al), whereas others have suggested that individual states may ultimately develop their own regulatory frameworks. Still others have proposed alternative certification frameworks outside the traditional regulatory process. 

Whatever path ultimately emerges, the UpDoc clearance brings these questions into sharper focus. 

ASCO AI in Oncology is published by Conexiant under a license arrangement with the American Society of Clinical Oncology, Inc. (ASCO^®). The ideas and opinions expressed in ASCO AI in Oncology do not necessarily reflect those of Conexiant or ASCO. For more information, see Policies.