In 1982, seven people died from cyanide-laced Tylenol in Chicago. Johnson & Johnson responded with full transparency, recalling all 31 million bottles—over $100 million of inventory—and launching the first tamper-proof packaging standards. Their leadership set a new bar for corporate responsibility and public safety. As AI transforms healthcare four decades later, federal policy is now trending away from such transparency.
Adam Mosseri, Head of Instagram, recently published an essay arguing that authenticity is becoming “infinitely reproducible.” His thesis goes to the heart of our digital age. When AI can generate photorealistic content indistinguishable from reality, the calculus of trust changes fundamentally. Deepfakes are growing more sophisticated. AI generates photographs and videos that we cannot distinguish from those captured by cameras. Social feeds fill with synthetic content mimicking reality with unsettling precision.
Mosseri argues that “rawness” and imperfection prove authenticity; blurry photos and shaky videos signal genuine human creation. However, technology commentator Shelly Palmer rejects this view, calling it an “adorable fantasy.” AI now convincingly fakes imperfection, making any aesthetic defense moot. Both observers agree on a crucial point: we must fingerprint authentic media, rather than endlessly trying to detect synthetic content. Even this approach faces a challenge. Almost every modern content production tool uses some form of AI. The line between acceptable AI tool use and excessive generation must be drawn on a case-by-case basis; a universal solution is impossible.
Industrywide Need for Human Guidance
Industries focused on entertainment, commerce, and social connection already wrestle with misinformation and the need for human guidance. These concerns become even more critical when patients’ lives are at stake. Healthcare AI hallucinations—outputs generated by statistical probabilities but ungrounded in clinical reality—pose much higher risks than a misleading social media post. A diagnostic AI system might identify nonexistent patterns, leading to false positives. A clinical documentation system could generate detailed but fictional patient histories. A treatment recommendation system might suggest therapies based on misinterpreted or nonexistent clinical evidence. Each scenario threatens the fundamental trust between patients and caregivers.
At this critical moment, federal policy retreats from the transparency and oversight that patient safety requires. The Trump administration has proposed eliminating requirements for health software vendors to disclose how AI tools are developed and tested. These “model cards”—similar to nutrition labels for AI—are vital. They help healthcare organizations understand the risks AI systems pose to patients. Without them, health systems must bear the full burden of vetting AI tools. This forces them to work harder to prove the technology’s trustworthiness.
Deregulation and Weakened Transparency
This deregulatory push extends beyond disclosure requirements. Executive Order 14179, signed in January 2025, weakened the Biden administration’s AI safety and risk management guidance and directed agencies to remove “barriers” to AI innovation. Supporters of these actions contend that reducing federal regulation accelerates technological progress and positions the United States as a leader in global AI development.
The July 2025 AI Action Plan established a policy of “minimally burdensome” regulation, explicitly prioritizing speed-to-market over safety guardrails, raising concerns about whether such approaches sufficiently mitigate patient risk. A December 2025 executive order went further, creating a federal task force to challenge state AI laws—including Colorado’s requirement for bias audits in healthcare AI and Utah’s mandate that hospitals disclose when AI tools are used in patient care—prompting critics to argue that removing these checks may undermine public trust and patient safety.
The administration presents this approach as essential for global competitiveness. However, healthcare cannot be equated to social media. Patients confer unconditional trust on their caregivers during moments of profound vulnerability. This trust must never be outsourced to algorithms or delegated to probabilistic models optimized for efficiency instead of safety.
EU Focus on Patient Safety
The European Union offers a contrasting model to recent U.S. policy shifts by explicitly classifying healthcare AI as high-risk because of its direct impact on health and safety. While the U.S. has recently reduced federal oversight—deprioritizing regulatory scrutiny and transparency in the name of innovation—the EU’s Artificial Intelligence Act imposes rigorous requirements for data quality, technical documentation, transparency, and mandatory human oversight.
For example, the Act requires that AI-driven medical diagnostic tools undergo strict pre-market conformity assessments and continuous post-market monitoring, placing patient safety and trust at the center of AI governance. This contrasts sharply with the current U.S. approach, which prioritizes speed-to-market and innovation, even if it means fewer regulatory safety guardrails.
The federal retreat from oversight makes organizational governance absolutely critical. In light of this shift, it is the direct responsibility of healthcare organizations to establish robust frameworks for AI verification—frameworks that federal standards should have supported. Organizational leaders must ensure the implementation of real-world performance monitoring, bias detection, and regular revalidation as essential duties, regardless of regulatory requirements. Drawing a parallel to Johnson & Johnson’s proactive decision to adopt transparency before tamper-proof packaging became compulsory, healthcare administrators and governing bodies are now obligated to prioritize safety over expedience to protect patient well-being.
AI Should Not Replace Human Judgment
The core principle must be that AI augments, not replaces, human clinical judgment. Healthcare providers represent a crucial source of insight into AI system performance through their daily interactions with these tools. They occupy the ideal position to identify problems, question recommendations, and apply clinical expertise when AI outputs seem implausible. Creating effective channels for collecting and analyzing this feedback transforms clinicians from passive AI consumers into active partners in maintaining system integrity.
Building trust demands more than technical training. It requires creating an environment where clinicians feel confident using AI content alongside their clinical judgment—knowing when to rely on AI recommendations and when to reject them based on expertise and patient-specific context. Patients deserve confidence that organizations use their health information appropriately and that AI-driven recommendations serve their best interests. This confidence emerges only from transparency about AI’s role in their care, clear explanations of data protection, and assurance that human oversight remains central to all clinical decisions.
The social media industry’s struggle with synthetic content offers a cautionary tale. As Mosseri observes, we are genetically predisposed to believing our eyes. Malcolm Gladwell articulated in Talking to Strangers that we default to truth because the evolutionary and social benefits of efficient communication outweigh the occasional cost of being deceived. This default to truth, so beneficial in most human interactions, becomes dangerous when sophisticated AI can exploit it. In healthcare, where patients already extend extraordinary trust to caregivers, the potential for AI-generated misinformation to cause harm multiplies dramatically.
Powerful Tool Controlled by Clinicians
The path forward requires treating AI as what it truly is: a powerful productivity tool that must remain under human guidance throughout the patient journey. From initial diagnosis through treatment planning and ongoing care management, clinicians must maintain meaningful oversight of AI recommendations. This is not a limitation on AI’s potential but rather the foundation for its safe and effective deployment.
When Johnson & Johnson faced its crisis in 1982, the company understood that the long-term value of public trust far outweighed the short-term costs of transparency. In the context of healthcare AI, embracing this same calculus is crucial. Therefore, policy recommendations must focus on several explicit actions: healthcare organizations should mandate transparent reporting of AI model development and validation, establish protocols for real-time monitoring and bias detection, and require continuous clinician involvement in reviewing AI output.
To implement these recommendations, organizations should adopt standardized reporting frameworks, such as model cards, for all deployed AI systems, integrate automated monitoring tools that flag anomalous outputs for clinical review, and create formal feedback processes that allow clinicians to routinely annotate and assess AI recommendations during patient care. Furthermore, periodic multidisciplinary audit meetings should be scheduled to review model performance and address any emerging issues.
The choices organizations make now will not only determine immediate patient outcomes but will also influence the broader societal trust in technology-driven healthcare. More broadly, the way healthcare organizations govern AI will establish vital precedents for balancing technological advancement and ethical responsibility in critical sectors. Ultimately, embedding transparency, oversight, and ethical responsibility into AI governance is essential to ensure that technological advancement serves the best interests of patients, shapes public expectations regarding the use of AI in sensitive domains, and upholds confidence in the healthcare system, even as federal requirements shift.
Our Duty of Care
Healthcare leaders must step up. Build trust through robust oversight, continuous monitoring, and a relentless focus on patient safety. Act now to balance innovation with responsibility; implement meaningful governance that ensures patient well-being, regardless of federal policy. Your leadership is essential—the future of healthcare AI depends on it.
Sources:
Mosseri, A. [@mosseri]. (2026, December 31). Looking forward to 2026, one significant interesting is that authenticity is becoming infinitely reproducible [Post on Threads]. Threads. https://www.threads.com/@mosseri/post/DS76UiklIDf (Accessed January 2026)
Palmer, S. (2026, January 2). The authenticity shortage. ShellyPalmer.com. https://shellypalmer.com/2026/01/the-authenticity-shortage/
Chaiken, B. P. (2024). Trust, Distrust, and Hallucinations (pp. 207-220) in Future Healthcare 2050: How Artificial Intelligence Transforms the Patient-Physician Journey. Poplar Tree Media.
0 Comments