What happened
Researchers from Sichuan University published a study in BMC Health Services Research applying the Policy Modeling Consistency (PMC) Index to 42 medical AI policy documents from the US, UK, EU, China, Japan, South Korea, and Australia. The analysis found that while 69% of policies received a Grade B rating, only 4.76% achieved the top Grade A. The study identified a structural imbalance where policies score high on strategic intent and perspective but low on specific policy content and regulatory tools necessary for execution.
A study led by Fei Wang and colleagues at Sichuan University analyzed 42 national-level medical AI policy documents issued between 2019 and 2025 by seven major jurisdictions: the United States, United Kingdom, Japan, South Korea, Australia, China, and the European Union. The research, published in BMC Health Services Research, utilized the Policy Modeling Consistency (PMC) Index, a quantitative framework designed to measure the internal coherence, structural completeness, and implementation readiness of policy texts.
The PMC methodology involved constructing a multidimensional evaluation system with nine primary variables and 35 secondary variables. These dimensions assessed aspects such as policy nature, purpose, content depth, tools deployed, and evaluation mechanisms. Each variable was scored to create a composite index and a concavity index, grading policies on a five-point scale from A to E. The results showed that only two policies (4.76%) earned a Grade A, while 29 (69.05%) received a Grade B, nine (14.51%) were Grade C, and two (4.76%) were Grade D. The overall mean PMC index was 6.67, interpreted as above-average but not exceptional quality.
The analysis revealed a distinct pattern: policies scored highest on dimensions related to policy nature, evaluation, and perspective, indicating strong strategic framing and broad intent. Conversely, the weakest scores appeared in policy content and policy tools, which are critical for execution. The authors describe this as a structural imbalance where comprehensive strategic frameworks are paired with weak implementation instruments, lacking concrete regulatory levers, funding mechanisms, and measurable milestones.
The study compared different governance approaches, noting that the US relies on agency-level guidance like FDA frameworks, the UK uses a principles-based regulator-led model, and the EU embeds health AI within horizontal legislation like the and GDPR. China, Japan, and South Korea have coupled industrial policy with health-sector pilots, while Australia has taken a standards-driven path. The PMC analysis allowed these diverse philosophies to be compared on a common quantitative footing for the first time at this scale.
Source details: bioengineer.org ↗
Why it matters
This research provides the first large-scale, quantitative comparison of how major economies are governing medical AI. It highlights a critical gap between high-level strategic ambitions and the operational reality of clinical deployment. By identifying that most policies lack concrete regulatory levers, funding mechanisms, and accountability structures, the study warns that the absence of specific rules may shift decision-making to courts and individual clinicians, potentially increasing risks related to patient safety, data privacy, and . The findings suggest that current policy frameworks are often aspirational rather than enforceable, which could hinder the safe and equitable integration of AI in healthcare systems globally.
The gap between strategic vision and operational enforcement has significant implications for patient safety and innovation. Medical AI touches sensitive areas such as clinical decision-making, data privacy, and liability. The study found that most policies exhibit robust strategic orientation but insufficient regulatory and ethical detailing. This absence of specific rules does not create a regulatory vacuum but shifts decisions to courts, hospitals, and clinicians, often after harm has occurred or trust has eroded.
The research highlights the risk of static policy documents being poorly matched to fast-moving technologies like and foundation models. A policy system that cannot revise itself quickly risks perpetually regulating the previous generation of technology. The authors suggest this is a global issue, not unique to any single country, and emphasize the need for dynamic policy adjustment systems.
The study demonstrates the value of treating policy text as data, using text mining and quantitative modeling to convert governmental prose into comparable scores. This approach makes cross-national policy assessment reproducible rather than subjective. In a domain where hype often outpaces evidence, this rigor provides a template for stress-testing drafts before publication, helping to identify weaknesses before they lead to implementation failures.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
Why can ethical evaluation not be reduced to one model score?
What to watch next
Policymakers may use the PMC Index as a diagnostic tool to stress-test new AI health regulations before publication. There is likely to be increased pressure on governments to update static policy documents to address dynamic technologies like and foundation models. Specific attention may turn to how jurisdictions like the US, EU, and China refine their data governance and ethical accountability mechanisms to close the identified implementation gap.
Policymakers may adopt the PMC Index as a diagnostic checklist for drafting new AI health regulations, ensuring documents specify responsibility, compliance monitoring, resources, and success metrics. This could lead to more operationally ready policies in upcoming legislative cycles.
There is likely to be increased focus on refining data governance mechanisms and strengthening ethical accountability, particularly in jurisdictions like China, where the study offers pointed recommendations. The emphasis on dynamic policy adjustment suggests that future regulations may include mechanisms for rapid revision in response to technological changes.
The findings may influence how international bodies and national governments approach the harmonization of AI standards in healthcare. The comparative nature of the study provides a baseline for assessing progress in closing the implementation gap across different regulatory philosophies.