Researchers and health organizations are weighing how AI could improve medicine without weakening evidence, accountability or human judgment. In a PLOS Digital Health opinion, Shenwei Wan and Xiaoguang Lei propose an AI-enabled “evidence commons” with secure data access, shared standards and stronger local research capacity. They note that a pivotal US drug trial had a median cost of $19 million, while 83% of 784,585 trial sites from 2006 to 2012 were in 25 high-income OECD countries. AI can help with literature searches, screening citations, translation and data extraction, but digital health tools still need monitoring and validation.
Other medical articles highlight limits and governance challenges. Pharmacology educators Joshua L. Hood and David W. Hein say integrated organ-based curricula can make core drug principles harder to teach and assess; 67% of accredited US MD-granting schools offered a separate pharmacology course 20 years ago, while 14% required more than one. Orthodontics researchers warn that inconsistent AI performance and complex patient-specific changes make automated predictions no substitute for clinicians. Wenke Liedtke and Martin Langanke argue that clinical AI accountability also requires shared moral standards, informed participation, meaningful alternatives and clear responsibility.
In clinical research, NetraMark co-founder Joe Geraci says its NetraAI can identify Phase II patient subgroups that may respond to a drug even when overall results are weak, helping sponsors decide whether to continue development and how to design later trials. He says placebo responders may also respond to the drug and stresses that AI recommendations must be transparent to researchers and regulators. OSF HealthCare, meanwhile, is exploring ambient listening, robotics and automation to reduce administrative work. CEO Michelle Conger says its governance process considers effects on patients and employees, ethics, labor, education and disinformation, with human dignity central to the approach.
AI’s risks and opportunities also feature in cybersecurity and copyright debates. Chinese cybersecurity firm Qi An Xin says it is responding to threats in the agentic AI era, as AI-enabled attacks target banks and governments. Calif used AI to help find Java4Shell, a vulnerability affecting enterprise applications and cloud infrastructure; a straightforward patch was deployed, and home users do not need to act. Separately, an Israeli American cloud security company used AI models to find serious flaws in PHP, followed by increased hacking attempts. On copyright, Australian creatives oppose weakening protections for AI training, while a review of cases involving Thomson Reuters, Anthropic, Meta and GitHub urges organizations to check data permissions, licenses and vendor reuse policies, and to preserve human review of consequential decisions.
AI policy and work remain contested. US President Donald Trump said he and Chinese President Xi Jinping support reasonable safety rules and will work together, while China’s embassy called for coordination, prevention of security risks and keeping AI under human control. The US and China are competing to develop advanced models; Trump said he did not discuss Nvidia or CEO Jensen Huang’s reported connection to a woman accused of spying for Beijing in depth, and Reuters said it had not independently confirmed the allegation. In New York, Gleb Tsipursky says falling entry-level white-collar postings make it harder for young people to start careers as employers adopt AI. Michele, a former lending quality analyst, argues AI should handle repetitive work while workers supply business-specific knowledge; she says an approved workplace tool sometimes gave six different answers to the same prompt and files.
Former Senator Phil Gramm, speaking at a Snow Institute lecture earlier in fall 2026, compared AI with the Industrial Revolution and argued it could make desired goods and services more affordable without producing the job losses many fear. He cited five million Americans quitting or losing jobs each month and an average of 5.1 million finding new jobs. Gramm advised students to pursue broad education and learn to work with people, arguing AI cannot lead, motivate or deal with others.
Key Takeaways
- A proposed AI-enabled medical evidence commons would combine secure access, shared standards and local research capacity; the article cites a $19 million median cost for a pivotal US drug trial and highly concentrated trial-site locations.
- Medical AI requires ongoing validation and human judgment, with researchers raising concerns about pharmacology education, orthodontic predictions and differing moral standards for clinical accountability.
- NetraMark’s NetraAI can identify potentially responsive patient subgroups in Phase II trials; Joe Geraci says recommendations should be transparent and placebo responders deserve examination.
- OSF HealthCare is exploring ambient listening, robotics and automation to reduce administrative work, with governance focused on human dignity and effects on patients and employees.
- AI is being used by cybersecurity defenders and attackers: Calif helped identify the Java4Shell vulnerability, while reports also describe AI-enabled attacks and increased hacking attempts after PHP flaws were found.
- Copyright rulings involving Thomson Reuters, Anthropic, Meta and GitHub prompt institutions to audit training and uploaded data, permissions, licensing and human oversight.
- Australian creative groups oppose weakening copyright protections for AI training, citing potential effects on rights holders worldwide.
- Trump says he and Xi support reasonable AI safety rules; China’s embassy calls for coordination, while the US and China compete to build advanced models. Trump’s remarks referenced Nvidia and Jensen Huang, but Reuters had not independently confirmed the reported spying allegation.
- New York’s declining entry-level white-collar postings raise concerns about young workers’ access to first jobs; a former analyst says AI should reduce repetitive work while people provide business context.
- Phil Gramm argues AI may create new jobs and make goods and services more affordable, and advises students to build broad skills and learn to work with people.
Researchers weigh AI’s role in widening access to medical evidence
In an opinion in PLOS Digital Health, Shenwei Wan of Renmin University of China and Xiaoguang Lei of the First Affiliated Hospital of Kunming Medical University examine whether AI could broaden access to medical evidence or create a new monopoly. They note that a pivotal US drug trial had a median cost of US$19 million, and research found that 83 percent of 784,585 trial sites were in 25 high-income OECD nations from 2006 to 2012. The authors propose an AI-enabled evidence commons with secure data access, clear methods, shared standards and stronger local research capacity. They also say changing digital health tools need continued monitoring and validation, alongside traditional trials, and describe AI tools that can help with literature searches, citation screening, translation and data extraction.
Compressed curricula and AI challenge pharmacology education
In a perspective in Frontiers in Pharmacology, Joshua L. Hood and David W. Hein of the University of Louisville School of Medicine examine challenges in teaching pharmacology to medical and dental students. Schools are moving from separate preclinical courses to integrated organ-based curricula, which can make it harder to teach core drug principles and check what students have learned. The authors say professional pharmacology education differs from graduate training, making it harder to find faculty prepared to teach clinical pharmacology. They report that 67 percent of accredited US MD-granting schools offered at least one separate pharmacology course 20 years ago, while 14 percent required more than one. Reports also indicate that students can struggle with pharmacokinetics when the subject is split across organ-based modules.
Review warns AI still has major gaps in orthodontic care
Muhammed Al-Huda Ballouk and Mohamed Altinawi review AI in healthcare, dentistry and orthodontics in an article in Philosophy, Ethics, and Humanities in Medicine. They point to inconsistent diagnostic performance, limits in what AI systems can know, and ethical concerns, warning that automated predictions should not replace clinical judgment. The authors say tools such as cone beam computed tomography and digital oral scanners have clear uses, but inaccurate algorithmic outputs can create risks when clinicians rely on them without question. They stress that orthodontic care depends on human expertise, communication and knowledge built through clinical experience. Treatment plans also change through ongoing discussions between doctors and patients, while complex facial growth and tissue responses are difficult to capture in databases.
Researchers say moral differences complicate responsibility for medical AI
Wenke Liedtke and Martin Langanke examine responsibility for machine learning systems used in clinical decision support in a conceptual analysis published in Philosophy, Ethics, and Humanities in Medicine. They say clinicians, patients and healthcare institutions may hold different moral and cultural values, making it difficult to set shared rules for decisions made with AI. The authors use a relational model that considers who is responsible, to whom, before which authority, for what action and under which standards. They argue that a shared moral standard is needed for responsibility to be meaningful, along with voluntariness, adequate information, self-determination with accountability and real alternatives. Without common standards, they warn, moral accountability may be reduced to legal liability or bureaucracy.
Trump and Xi support moderate AI safety rules, Trump says
US President Donald Trump said in a Fox News interview on Thursday, November 19, that he and Chinese President Xi Jinping want reasonable AI rules for safety and will work together on them. The leaders met in Shenzhen, China, during Trump’s visit. China’s embassy in Washington said the countries should improve coordination, prevent AI security risks and keep AI under human control. The US and China are competing to build advanced AI models, which both see as important for military and technology leadership. Trump said he did not discuss Nvidia or CEO Jensen Huang’s reported relationship with a woman accused of spying for Beijing in depth; Reuters said it had not independently confirmed that the woman was a spy.
Qi An Xin uses AI to defend against rising cyberattacks
Qi An Xin Technology Group vice president Chen Huaping spoke at a cybersecurity conference in Xiamen, Fujian, last week. The article profiles the Chinese cybersecurity firm as it responds to new risks in the agentic AI era. It reports that AI enabled cyberattacks are affecting banks and governments worldwide. The story also notes that an Israeli American cloud security company used new AI models to find serious flaws in PHP, which is used by an estimated 70 per cent of websites, followed by a rise in hacking attempts.
New York’s AI boom needs a plan to help young workers get started
Gleb Tsipursky argues that New York needs a commitment to help young people find their first jobs as AI changes the state’s economy. Demand for skilled workers is growing faster than supply, while entry level job postings in white collar fields have fallen sharply. Employers increasingly use AI for tasks such as automation, data analysis and decision making. The article says this makes it harder for young people to begin their careers.
AI can handle routine tasks while workers bring their experience
Michele, a former home lending quality assurance analyst, lost her role in a corporate restructuring earlier this year and now uses AI in her own business. She argues that companies should give workers AI to reduce repetitive tasks, not use it to replace them. At her former job, the approved tool sometimes gave six different answers to the same prompt and files. Michele says AI works best when it has details about a business’s rules and exceptions, knowledge that comes from people’s experience. Without that context, AI may provide a generic answer or make something up.
Australian creatives warn against weakening copyright rules for AI training
Australia’s creative sector is concerned that AI companies may persuade the government to weaken copyright protections for training AI models. The music industry says such a change could affect rights holders around the world. The article places the debate within Australia’s wider arguments over AI policy and copyright.
Copyright cases push institutions to review AI data and oversight
Recent copyright rulings show why employers, universities, nonprofits and public agencies should check permissions before using material in AI systems. Courts reached different conclusions in Thomson Reuters v. ROSS Intelligence, Bartz v. Anthropic, Kadrey v. Meta and Doe v. GitHub, including concerns about pirated copies and AI generated code. Institutions should review ownership, where copies came from, license rules, confidentiality duties and whether vendors retain or reuse uploaded files. The article also urges organizations to identify decisions AI affects, assign approval roles and ensure people can review and override recommendations in important cases. It cites guidance from the National Institute of Standards and Technology and human oversight requirements in the EU AI Act.
OSF HealthCare puts human dignity at the center of its AI plans
OSF HealthCare CEO Michelle Conger said Pope Leo XIV’s first encyclical on artificial intelligence offers Catholic health care organizations guidance for using AI while protecting human dignity. Speaking at the OSF OnCall Digital Health Symposium, she said technology should serve patients, communities and health care workers. OSF is exploring tools such as ambient listening, robotics and automation to reduce administrative work and help employees focus more on patient care. Its governance process considers effects on patients and employees, along with ethics, labor, education and the risk of disinformation.
AI helps find Java security flaw that could threaten major systems
Cybersecurity firm Calif used artificial intelligence to help find a vulnerability in Java, a programming language used by many large business, finance and government applications. The flaw, called Java4Shell, was in the environment that runs company code, so attackers could potentially have crashed vital systems or gained remote access to valuable information. It affected enterprise applications and cloud infrastructure, not home devices. Iowa State University cybersecurity researcher Doug Jacobson said experts deployed a straightforward patch, and home computer users do not need to take action. The discovery shows how both hackers and defenders can use AI to analyze large amounts of code for weaknesses.
NetraMark’s Joe Geraci explains how AI can uncover drug trial signals
In an interview with Pharmaceutical Executive, NetraMark co-founder and CSO/CTO Dr. Joe Geraci discussed how the company’s NetraAI analyzes clinical trial data. The system can identify patient subgroups in Phase II trials where a drug may work, even when the overall results are too weak to support a business decision. Geraci said this analysis can help sponsors decide whether to pursue an asset in later trials and design those trials around likely responders. He also said placebo responders may also respond to a drug, and researchers should examine that relationship rather than treating the groups as separate. Geraci emphasized that AI recommendations in clinical trials must be transparent so researchers and regulators can understand how a model reached them.
Phil Gramm says AI may bring new jobs and affordable goods
Former Senator Phil Gramm compared artificial intelligence with the Industrial Revolution during a Pope Lecture hosted by the Snow Institute for the Study of Capitalism earlier in fall 2026. He said the Industrial Revolution raised living standards for 65 million people and argued that fears of widespread unemployment did not match what happened. Gramm expects AI to help provide goods and services that people want but cannot currently afford, and said job loss fears differ from labor statistics. He noted that five million Americans quit or lose jobs each month, while an average of 5.1 million find new jobs. He advised students to get a broad education and learn to work with people, saying AI will not lead, motivate or deal with others.
Sources
- Researchers Examine How Artificial Intelligence Could Reshape Clinical Evidence Production Systems
- Medical Pharmacology Education Faces Strains From Compressed Curricula and Artificial Intelligence
- Epistemological Review Finds Artificial Intelligence Shows Critical Gaps in Orthodontic Care
- Moral Diversity in Clinical Care Complicates Ethical Responsibility for Medical Artificial Intelligence
- Trump Says He and China's Xi Want Moderate AI Rules
- Meet China’s cybersecurity giant fighting AI attacks with AI
- Op-ed: New York’s AI boom needs a first job commitment
- AI Is the Side Dish. Your Experience Is the Main Course.
- Australian creative sector bristles at ‘permit’ AI training plans
- AI Copyright Rulings Put Data Governance on the Institutional Agenda
- OSF using Pope's encyclical to guide AI strategy
- AI helps uncover critical flaw in widely used internet software amid cyber arms race
- How AI Is Rescuing Failed Clinical Trials: Q&A With Dr. Joe Geraci
- Former Sen. Phil Gramm discusses AI, jobs and economic change
Comments
Please log in to post a comment.