Murielle Beene, chief health informatics officer at Trinity Health, discusses the evolving role of AI in clinical informatics. Beene emphasizes the importance of integrating AI tools thoughtfully into health care workflows to enhance patient and provider experiences. She highlights the challenges of rapidly changing technology and the need for clear communication about AI's capabilities to avoid unrealistic expectations. Trinity Health, which encompasses 92 hospitals across 25 states, is focusing on AI applications that improve operational efficiency while maintaining patient safety and ethical standards.

Researchers have developed an AI-based endoscopic scoring system that better aligns with symptomatic measures and quality of life in ulcerative colitis patients than the conventional Mayo endoscopic score. The study, involving 387 patients and published in the American Journal of Gastroenterology, found that the AI-generated cumulative disease score provided a more accurate reflection of therapeutic benefits, particularly for the drug ozanimod. However, the AI system does not capture all symptoms of inflammatory bowel disease, indicating areas for further improvement.

Researchers have developed an AI model to identify patients with congenital heart disease who are at risk of experiencing gaps in continuous care, as presented at the American Heart Association Scientific Sessions. The model, which utilizes data from 1,746 patients, takes into account factors such as social determinants of health and disease complexity. While the model shows strong performance in identifying patients maintaining follow-up, it needs refinement to better predict those with care gaps.

The FDA has qualified AIM-NASH, the first AI tool designed to assist in assessing severe fatty liver disease during drug trials. This cloud-based system analyzes liver tissue images to identify signs of diseases like MASH, potentially streamlining clinical trials by standardizing assessments and reducing time and resources. The FDA's decision was based on studies showing that AI-assisted results are comparable to those of expert evaluations, underscoring the growing role of AI in drug development.

Experts at the Food and Drug Law Institute conference have urged the FDA to provide clearer guidance on the regulation of AI and large language models used in drug development and manufacturing. Anthony Schiavone from Agilent Technologies highlighted the fragmented global regulatory landscape, noting the EU's comprehensive AI framework. Tala Fakhouri from Parexel shared that over 800 AI-related submissions have been made to the FDA's Center for Drug Evaluation and Research, mostly in oncology.

A six-month investigation has revealed over thirty security vulnerabilities in AI-assisted development tools, including popular IDEs like Visual Studio Code and JetBrains products. The IDEsaster report shows how these tools can be manipulated for data exfiltration and remote code execution. Security researcher Ari Marzouk said that current IDEs fail to incorporate the "Secure for AI" principle, necessitating a fundamental redesign to address these flaws.

Gartner advises organizations to block users from downloading or installing AI browsers like Perplexity's Comet and OpenAI's ChatGPT Atlas due to security risks. These browsers, which feature AI sidebars and agentic transaction capabilities, prioritize user experience over security, potentially exposing sensitive data. Gartner highlights the risk of data exposure and unauthorized actions if these browsers are deceived. The firm recommends conducting risk assessments and educating users about the potential for data to be sent to AI back ends.

About 1 million children in the US developed obesity during the COVID-19 pandemic, increasing the obesity rate from about 21% to 23%, according to a study in the journal Pediatrics. Researchers found that minority and disadvantaged youths, as well as those already overweight, were particularly affected.