JMIR AI

No preview text available.

No preview text available.

AI has increasingly been used to generate synthetic images to overcome the scarcity of large, regulatory-compliant medical image datasets. However, their quality and actual utility for algorithm development across deep learning architectures, medical settings, and image types have been poorly assessed.

The proliferation of evaluative approaches for generative AI tools used for mental health purposes reflects much-needed attention to safety, ethics, and emerging risks. However, limited alignment across frameworks complicates comparisons, slows progress toward shared goals, and provides little clarity for regulatory processes. The result is a fragmented field that lacks cohesive guidance as the technology and its real-world patterns of use continue to evolve. This position paper synthesizes current evaluative approaches, outlines sources of divergence across existing frameworks, and identifies shared domains that could serve as anchors for coordinated development. We advocate for a living, flexible meta-framework maintained by a global, cross-sector consortium that can harmonize criteria, support consistent evaluation across jurisdictions, and adapt alongside scientific, clinical, and technological advances. Such a structure can strengthen safety, transparency, and methodological consistency while giving researchers, clinicians, developers, and policymakers a shared foundation for assessing generative AI tools in mental health care.

Measles is a major cause of death among infants despite free routine immunization (RI) in low- and middle-income countries (LMICs). Vaccination defaulting contributes to measles outbreak occurrences. Conventional immunization defaulter identification methods are expensive to implement, prone to data falsification, reactive, and overestimate uptake. Several machine learning studies leveraged the use of child and maternal characteristics as predictors of immunization defaulting, overlooking community, demographic, and socioeconomic factors. Similarly, there is limited research regarding the use of interpretable deep learning and advanced ensemble methods to improve model performance in predicting RI defaulters and establishing significant predictors at the national level in the LMIC context, like Uganda.

Large language model–based chatbots (LLM-CBs) are increasingly used as mental health support tools. Risks and harms are discussed especially for unsupervised use and for vulnerable groups. A naturalistic characterization of the use of LLM-CBs by patients with mental health disorders is currently lacking.

Adverse event reports referring to the same case but treated as independent can negatively impact statistical analysis and may mislead clinical assessment. Pharmacovigilance relies on large databases of adverse event reports to discover potential new causal associations. Their size necessitates computational methods to identify duplicates at scale. The current state-of-the-art is statistical record linkage which outperforms rule-based approaches. In particular, vigiMatch is in routine use for VigiBase, the World Health Organization global database of adverse event reports, and represents the first statistical duplicate detection approach in pharmacovigilance deployed at scale. Originally developed for both drugs and vaccines, its application to vaccines has been limited due to inconsistent performance across countries.

AI technologies are currently experiencing significant growth in development and use in health care. The AI model lifecycle includes several stages, all of which should be adequately described. Reporting standards are commonly used for this purpose. However, the plain-text structure of these standards prevents transparent machine-readability of this information.

Symptom detection is essential in global disease surveillance to detect potential outbreaks, as symptoms are the first observable signs of infection. To reflect real-time ground truth conditions during pandemics, social media has emerged as a valuable data source. Moreover, effective digital disease surveillance systems must operate across diverse linguistic settings, and large language models (LLMs) have been shown to perform inconsistently across languages, tending to have lower performance in low-resource languages. While multilingual approaches have been explored in various health-related natural language processing tasks, a critical gap remains in understanding whether LLM-based symptom detection can perform consistently across languages for global disease surveillance. Southeast Asia demonstrates this challenge, combining diverse languages and the potential for emerging infectious disease outbreaks, making it a case for evaluating how multilingual performance disparities manifest in symptom detection.

Independent audits of medical large language models have concentrated on models that answer unsafely rather than those that decline safe questions. In June 2026, Anthropic released Claude Fable 5 with a safeguard that detects requests related to cybersecurity, biology, or chemistry and reroutes them to a fallback model, Claude Opus 4.8. The developer states that this safeguard is deliberately conservative and affects fewer than 5% of sessions. Whether that figure holds for consumer health questions, many containing dense biomedical vocabulary, was unknown.

Personalizing physical activity recommendations for older adults requires understanding not only which dimensions of physical activity and sedentary behaviors (24-h movement behaviors) influence health outcomes but also when, within an individual’s everyday life, these dimensions are most relevant. Current observational and interventional approaches rarely capture the temporal dynamics linking everyday patterns of 24-hour movement behaviors to cognitive and mental health trajectories, 2 key determinants of healthy aging.

Conversational AI systems are increasingly being deployed in health care for clinical decision support, but their performance varies substantially across patient communication styles, health literacy levels, and behavioral patterns. Static benchmarks cannot capture multiturn dynamics through which this variation compounds, and no current evaluation framework implements structured AI risk management guidance for conversational health care AI. The result is a structural risk: AI systems may perform well in aggregate while failing disproportionately for the populations they are intended to help.