The recent US government shutdown has left critical federal health websites and datasets outdated, posing risks as the flu season approaches. Key resources like the CDC's Morbidity and Mortality Weekly Report, wastewater surveillance data used to track the spread of COVID, flu and RSV, and COVID-19 vaccine coverage data remain not updated, raising concerns among public health experts about the nation's preparedness. While some data, such as FDA drug recalls, have been maintained, the delay in other updates highlights the importance of federal data tracking for timely public health responses.

"I'm worried we had a monthlong blackout of flu data at a time when it's flu season and they're not talking about flu vaccines at all," said Jennifer Nuzzo, a professor and director of the Pandemic Center at Brown University School of Public Health.

At AMP 2025, bioethicist Takunda Matose, PhD, discussed the ethical considerations of using AI in health care. He emphasized the importance of mindful data usage, addressing inherent biases, and the need for human discretion in AI systems. Matose argued that while AI excels in data analysis, its limitations due to bias and ethical challenges must be acknowledged. He urged stakeholders to consider the broader implications of AI, focusing on patient and stakeholder needs, and to thoughtfully integrate AI into health care systems.

OpenAI's latest large language model, OpenAI o3, demonstrated its potential in educational settings by scoring 90% on Japan's licensing exam for radiologic technologists and generating mock exam items that met national standards. The study, led by Toshimune Ito from Teikyo University, highlighted the model's strengths in item difficulty, factual accuracy and content coverage, though it noted areas for improvement in wording clarity and instructional usefulness. Future advancements in visual encoders and adaptive feedback could enhance its application in radiological technology education and beyond.

Oded Rechavi, a molecular biologist at Tel Aviv University, developed an AI tool named q.e.d to enhance the peer review process for scientific manuscripts. The tool, which provides constructive feedback by breaking down research into a "claim tree," has been positively received by researchers for its ability to identify strengths and weaknesses in manuscripts. However, some, like immunologist and evolutionary biologist Mark Hanson, criticize it for lacking original insights. Since its launch, q.e.d has been used by researchers globally, with plans for future enhancements and collaborations.

Google DeepMind, co-founded by Demis Hassabis, has made significant strides in AI, notably winning the 2024 Nobel Prize in Chemistry for its AlphaFold model, which revolutionized protein structure prediction. The company, acquired by Google in 2014, aims to replicate this success across various scientific fields, including weather forecasting and nuclear fusion, despite the challenges posed by the rapid commercialization of AI and increased competition. DeepMind continues to focus on ethical AI development, balancing scientific innovation with responsible practices.

The Fleming Initiative and GSK are collaborating on a nearly $60 million AI-driven project aimed at combating drug-resistant infections in the UK. This initiative targets the discovery of new antibiotics and strategies to fight Gram-negative bacteria, which are resistant to many treatments. Dr. Andrew Edwards from Imperial College London is leading efforts to use AI to understand and overcome bacterial defenses. The project also aims to predict patterns of superbugs and address fungal infections, underscoring the urgent need for innovative solutions to a growing global health threat.

The concept of "agentic AI" is gaining traction in the tech industry, promising AI systems that can autonomously plan, act and learn, distinguishing them from traditional chatbots. Companies like Amazon, Google and OpenAI are exploring these AI agents, which could handle complex tasks like managing emails or making purchases. However, the term "agentic" is seen by some as marketing jargon, despite its potential to transform AI applications. Researchers from Massachusetts Institute of Technology and the Boston Consulting Group point out that these systems behave more like autonomous teammates, capable of executing multistep processes.

Microsoft has issued a warning regarding the new agentic AI capabilities coming to Windows 11, underscoring potential security risks. The company advises users to enable this feature only if they understand the security implications, as it will be off by default. This move is part of Microsoft's broader vision to evolve Windows 11 into an agentic operating system, despite user concerns about security vulnerabilities, such as the Xpia malware.

Significant disparities exist in vaccination rates among different income groups, with only 62.3% of children from low-income families receiving all early childhood vaccines, compared with 77.9% of high-income children. The Commonwealth Fund report highlights logistical barriers, such as transportation and scheduling issues, contributing to these disparities.