AI Health

Friday Roundup

The AI Health Friday Roundup highlights the week’s news and publications related to artificial intelligence, data science, public health, and clinical research.

June 12, 2026

In this week’s Duke AI Health Friday Roundup: new contender enters the protein-prediction field; clinical robot retired after annoying nurses; glucagon emerges as a factor in next-gen weight loss drugs; improving therapeutic ROI with predictive AI; hack compromises Github repositories; some scientists feel reluctance, FOMO when using AI; MLA Task Force releases framework for working with AI in humanities research, teaching; much more:

AI, STATISTICS & DATA SCIENCE

“The predictions were made using ESMFold2, an AI model that Biohub says surpasses the performance of AlphaFold3, the latest version of Google DeepMind’s system, and other protein-structure prediction AIs. The atlas is described in a report released today….Other scientists are impressed with the results, especially that ESMFold2 is fully open source. But the Biohub model enters an increasingly crowded field, in which competing open-source and proprietary protein models are making gains at breakneck speed.” Nature’s Ewen Callaway and Miryam Naddaf report on a new open-source protein-predicting model that exceeds the hundreds of millions of structures and sequences predicted by the previous champ, AlphaFold.
“Right away, Lapkin, an ICU nurse, noticed Moxi was in trouble. It frequently lost its way in hospital corridors, cried for help when faced with elevator buttons, and ping-ponged before elevator doors, unable to discern which lift to take. If something blocked its charging spot, Moxi spun in circles. Eventually, Lapkin said, Moxi needed a handler to accompany it and help it navigate.” Robots as adjuncts for nurses in the clinic? The experiences related in Varsha Bansal’s article for Proof News suggest there may be a ways to go before that vision can be realized successfully.
“Predictive AI is capable of improving treatment allocation functions as a multiplier on the ROI of the therapies it guides. By shifting the baseline risk of treated populations through risk-based targeting and increasing average treatment responsiveness through response-based targeting, predictive allocation can reduce effective NNT [number needed to treat] and increase average treatment effects, improving the realized ROI of existing therapies and strengthening incentives to develop new ones.” An article published in NEJM AI by Jonathan D. Ketchum makes a case for predictive AI as a means for improving treatment allocation and return on investment for therapies.
“Microsoft has shut down a wave of its own repositories on GitHub, including those related to Azure and AI coding agents, as it investigates a data breach, according to research from cybersecurity researchers and a statement given to 404 Media by Microsoft. Hackers planted malware that would harvest peoples’ credentials when they opened it in AI coding tools like Claude Code or Gemini CLI, according to one set of researchers.” At 404, Joseph Cox reports on a hack that led to a large-scale shutdown of Microsoft Github repositories after hackers succeeded in sneaking credential-harvesting malware into a repository commit.

BASIC SCIENCE, CLINICAL RESEARCH & PUBLIC HEALTH

“In the new study, which was published on Sunday in the journal Frontiers in Ocean Sustainability, researchers trained an AI algorithm on hundreds of three-dimensional x-ray images—the kind of imaging already used in airports—of 68 dried shark fin, seahorse and sea cucumber samples. Across hundreds of images, the algorithm correctly identified these samples 92 percent of the time, with a false positive rate of about 13 percent.” Thinking of smuggling a squid in your luggage? OK, probably not, but if you did, AI might be able to tell, according an article in Scientific American by Jackie Flynn Morgensen.
“Perhaps most intriguing is glucagon’s potential to counter one of weight loss’s biggest biological hurdles: the body’s tendency to conserve energy as pounds drop. As people lose weight, metabolism typically slows. The body begins burning fewer calories, making further weight loss harder and increasing the likelihood of regaining weight. Human studies examining glucagon’s effects on calorie burning have produced mixed results. But Campbell presented new animal data suggesting glucagon receptor activation may help prevent that metabolic slowdown.” The Duke School of Medicine’s Shantell Kirkendoll reports on the surprising role being played by the hormone glucagon in the development of next-gen weight loss drugs.
“Of the 1,907 scientists who responded, nearly 48% said that they feel negative towards AI. Meanwhile, 30% of respondents to the poll said that they feel positive about AI overall, and 22% were neutral…Only 23% of the respondents felt that AI tools were having a positive impact on research, whereas 31% said that the technology was negatively affecting science. Nearly half of the researchers felt that the impact of AI use depends on how the tools were used.” An international survey of scientists’ attitudes toward AI by Nature reveals a combination of skepticism, some acknowledgment of specific benefits, and a general fear of missing out.

COMMUNICATIONS & Policy

“This framework is designed to help the profession navigate new technical and social conditions. We encourage faculty members, administrators, editors, and publishers to use it as common ground for dialogue, consensus-building, and policy creation… This framework emerges from three interrelated areas—AI literacy, human-centered expertise, and responsibility and ethics—and is organized into thematic questions to guide discussion and critical reflection.” An MLA task force has released a framework for working with AI in humanities scholarship and teaching.
“The stakes are high. Communications must come to be seen as a global public good. Expertise and excellence in health communications are integral to advancing population health and well-being; positive physical and mental health outcomes; longevity and quality of life; and access and affordability across the world. Few would disagree. Many are concerned. And this is the problem: there is an information issue, but it is not clear how to solve it.” In a commentary article published in Nature Health, Stern and Mukherjee describe some of the challenges confronting efforts to communicate accurately, clearly, and effectively about health-related issues in the current climate.
“Under the current rules to avoid bias, reviewers can generally participate in a panel—and agency staff can help run the review—as long as they recuse themselves from decisions on proposals from their own institution….The new policy, laid out in an internal document obtained by Science, will go into effect on 3 August and prohibit both groups from participating in any panel if they have an institutional COI with any proposal.” Science’s Daniel Garisto reports on new, more stringent conflict of interest rules for volunteer reviewers at the National Science Foundation that may put an additional squeeze on workloads.
“In fiscal year 2025, the number of NIH-funded investigators declined for the first time in a decade, with the number of research grant PIs and fellowship recipients falling a relative 4.0% and 9.9%, respectively. These declines were not evenly distributed: investigators from groups historically underrepresented in science experienced disproportionately larger declines than their peers.” A research letter published in JAMA by Nguyen and colleagues examines recent trends in research funding awards from the National Institute of Health.