Theranos to AI: Why FDA’s Lagging Oversight Demands a Reboot

The specter of Theranos, a cautionary tale of unchecked innovation and devastating consequences, continues to loom large over the healthcare technology landscape. As artificial intelligence rapidly integrates into diagnostics and clinical decision-making, a critical question emerges for regulatory bodies and investors alike: how can we prevent an “AI Theranos,” where advanced algorithms, rather than flawed blood analyzers, lead to widespread patient harm? The FDA’s delayed intervention in the Theranos saga offers profound lessons for the proactive oversight of AI safety.

The Theranos Precedent: A Failure of Proactive Oversight

Theranos promised a revolution in blood testing, claiming its proprietary “Edison” device could perform hundreds of tests from a single drop of blood. The reality, as later exposed, was a sophisticated deception built on unvalidated technology and misleading claims. While the FDA eventually issued a warning letter to Theranos, this intervention came significantly late, after years of operation and widespread use of unproven diagnostics. The consequences were severe, impacting patient care and eroding public trust in novel medical technologies. The parallels to the burgeoning field of AI in healthcare are striking. Just as Theranos operated with a veneer of scientific advancement, many AI solutions today are presented with claims of unprecedented accuracy and efficiency. The challenge for regulatory bodies, particularly the FDA CDRH (Center for Devices and Radiological Health), is to differentiate between genuine innovation backed by robust clinical validation and unproven technologies that could pose significant health risks. Bakul Patel, who was a key figure in the FDA’s digital health efforts, has consistently emphasized the need for a balanced approach to innovation and safety. However, the Theranos case underscores that even with astute regulatory minds, the system can be slow to react to rapidly evolving, often opaque, technologies. The FDA’s delayed intervention on Theranos demonstrates the need for proactive AI safety oversight. Had stronger, earlier regulatory mechanisms been in place, or had the FDA CDRH been equipped with more agile enforcement capabilities, the extent of the harm might have been mitigated. Scott Gottlieb, former FDA Commissioner, has also spoken extensively on the need for modernizing regulatory frameworks to keep pace with technological advancements, a sentiment that resonates deeply when reflecting on the Theranos debacle. The narrative here is not about stifling innovation, but about ensuring that innovation is responsibly developed and rigorously validated before it reaches patients.

The AI Healthcare Landscape: Echoes of Unchecked Claims

Today, AI health misinformation news frequently highlights instances of AI chatbots providing incorrect drug interaction guidance or AI systems missing diagnoses. These failures, while perhaps not yet on the scale of Theranos, represent the early warning signs of a system that could be vulnerable to similar foundational flaws. Eric Topol, a vocal proponent of digital medicine, has also consistently called for rigorous validation of AI in healthcare, stressing that “AI must be held to the same, if not higher, standards of evidence as any new drug or device.” The core issue is often the lack of transparency and rigorous clinical validation, mirroring Theranos’s opaque operations. Companies developing AI solutions for health often tout impressive performance metrics derived from internal datasets, but these rarely translate directly to real-world clinical effectiveness or safety. The risk of undertriage of cardiac emergencies or delayed stroke identification by an unvalidated AI system carries profound implications for patient outcomes. The FDA CDRH, in its role of safeguarding public health, faces the immense task of scrutinizing these claims and ensuring that AI algorithms are not deployed prematurely. The challenge is compounded by the rapid pace of AI development. Unlike traditional medical devices, AI models can evolve, learn, and even “drift” in performance over time, necessitating continuous monitoring and re-validation. This dynamic nature demands a regulatory approach that is both stringent and adaptive, moving beyond static pre-market approvals to embrace a lifecycle management perspective for AI.

Regulatory Frameworks and the Path Forward

The existing regulatory landscape offers some foundational tools. The FDA SaMD Framework (Software as a Medical Device) provides guidance for software intended for medical purposes that operates independently of hardware. Many AI health solutions fall under this classification. Furthermore, the FDA 510(k) Pathway is a common route for medical devices demonstrating substantial equivalence to a predicate device. However, these pathways, while essential, may not be sufficiently robust to address the unique challenges posed by adaptive AI algorithms, especially those that learn and change post-market. To address this, the FDA finalized guidance on Predetermined Change Control Plans (PCCPs) in August 2025, providing a framework for managing modifications to AI/ML-enabled devices across their total product lifecycle. The Theranos case highlights the critical need for enforcement. The DOJ (Department of Justice) eventually stepped in, but only after significant patient exposure to the flawed technology. For AI in healthcare, the FDA CDRH must be empowered to act decisively and proactively. This includes not just issuing warning letters but also having clear mechanisms for auditing AI models, requiring transparent reporting of performance metrics, and mandating post-market surveillance that can detect algorithmic drift or emergent biases. FDA guidance on AI/ML medical device change control

A Mandate for Proactive AI Safety Oversight

The lessons from Theranos are unequivocal: delayed intervention can have catastrophic consequences. For AI in healthcare, the stakes are equally high. Regulatory officers at the FDA and investors in the healthcare AI space must recognize that the proactive oversight of AI safety is not merely a bureaucratic hurdle but a fundamental safeguard for public health and a prerequisite for sustainable innovation. The FDA’s delayed intervention on Theranos demonstrates the need for proactive AI safety oversight, a principle that must be rigorously applied to AI. The path forward demands a collaborative effort. Investors must conduct thorough due diligence, demanding robust clinical validation and clear regulatory strategies from AI companies. Regulators, particularly the FDA CDRH, must evolve their frameworks to be more agile, proactive, and capable of overseeing the dynamic nature of AI. This includes fostering greater transparency from developers, establishing clear benchmarks for real-world performance, and ensuring that enforcement mechanisms are swift and effective. Only by learning from the past and embracing a culture of rigorous, proactive AI safety oversight can we harness the transformative potential of AI in healthcare while simultaneously protecting patients from the next “AI Theranos.” Academic paper on ethical AI in healthcare Report on AI health misinformation trends