Is Anyone Training the Trainers?
Four 2026 papers move past what students should learn about AI and ask whether the faculty teaching them are actually ready.
This Week in MedEd
7 open funding calls, 24 upcoming events, 27 live jobs (15 new, 5 closing this week), including:
💰 EdTech Accelerator (up to €2.7M for European EdTech startups and SMEs, deadline 1 Oct 2026)
🎓 IAMSE 2026 Virtual Forum (online, global, 20-22 Oct 2026)
💼 Professor: AI & Health (University of Johannesburg, closes 9 Aug)
The full list, with deadlines, links and eligibility, is one tap away. Try everything free for 30 days.
Most of what gets written about AI in medical education assumes the hard part is getting students to use these tools responsibly and safely. But before we get there, somebody has to work out what a well-prepared faculty member actually looks like, and how you build one of those! This week’s papers step back from the student-facing question and look at the infrastructure behind it: the training programmes, competency scales, and consensus frameworks institutions are building so the people doing the teaching, marking, and mentoring are ready themselves.
I’m covering for papers this week, all providing advice on how we might ‘do AI’ next semester. Highlights include: evaluation of a faculty training programme, a validated scale for measuring what “AI-competent” faculty actually looks like , an international consensus framework for responsible AI practice, and a practical model for how we might not need to reinvent the wheel.
I was pretty down on the proliferation of endless frameworks and opinion pieces a few weeks ago… but I think these ones are useful.
Key points
A blended workshop programme in Pakistan produced a 22% knowledge gain among 290 faculty, but 3 months later fewer than one third had fully applied what they learned.
A new 16-item scale, validated across 434 clinical, technology and nursing staff in China, gives institutions a way to measure GenAI-integrated teaching competency.
A 303-participant international Delphi study distilled AI guidance for health professions education into ten consensus principles, reaching 96% final agreement.
A practical framework built for clinical teachers argues the real skill to teach students isn’t AI use itself, it’s verification, critical appraisal and ethical reflection around AI outputs.
None of these four papers is really about the technology. They are about the scaffolding institutions need to build around it.
Naeem et al evaluated the "New Knowledge Help" (NKH) programme, a blended series of six in person and two virtual workshops run across three Pakistani cities between November 2023 and March 20241. NKH aimed to help 290 faculty members to engage with AI ethically. They measured this using Kirkpatrick's four-level framework, and found strong satisfaction (mean 4.40-4.95 out of 5) and a 22% gain in post-workshop knowledge scores.
However three months after their intervention only 30% of participants reported fully applying NKH-aligned practices. The conclusion is that a good workshop can shift knowledge quickly, but translating that into practice needs ongoing institutional support.
Zhu-ge et al tackle a more basic problem, which is that there has been no standardised way to measure whether a faculty member is actually competent to teach with GenAI in a simulation setting2. Working from the TPACK (Technological Pedagogical Content Knowledge) framework, they ran a two-round Delphi process with 434 participants across clinical medicine, medical technology and nursing to build and validate a 16-item competency scale.
The resulting instrument showed excellent internal consistency (Cronbach's alpha 0.979) and an acceptable model fit, giving institutions a genuinely usable tool for auditing GenAI teaching competency rather than assuming it exists.
The GenAI-specific competencies are:
GenAI Tool Proficiency: Skilled use of generative AI tools (e.g., ChatGPT, DeepSeek, DALL-E, medical diagnostic AI) for lesson preparation and case
generation.
GenAI Content Verification: Ability to evaluate accuracy and reliability of AI-generated medical information.
GenAI-Driven Curriculum Design: Leveraging AI to analyze student data and tailor personalized teaching plans.
Knowledge of ethical guidelines (e.g., data privacy, AI diagnostic
accountability) for medical AI applications.
Training students to use GenAI tools for knowledge acquisition and case analysis with critical thinking.
Interpreting AI-generated teaching data (e.g., student knowledge/skill
metrics) to evaluate outcomes.
GenAI-Teaching Integration Innovation: Pioneering novel methods (e.g., AI-enhanced simulation cases) to diversify pedagogy.
Sonnenberg et al ran a modified Delphi process to build the Health CARE-AI framework (Contextual, Accountable, Responsible, and Equitable Artificial Intelligence ) to clearly articulate what responsible AI practice should look like across education, research and clinical care3.
303 participants took part across the three phases and 58 of 61 draft statements met the inclusion threshold before a final voting round confirmed all ten resulting principles. Endorsement was close to unanimous (96%).
The framework is:
Responsible artificial intelligence (AI) use is both an individual and collective duty. This clarifies responsibilities at individual, team, and institutional levels; sets expectations for reviewing tools; and keeps patients and learners at the center.
Use AI with honesty and integrity. Be transparent about AI assistance, avoid deception or manipulation, and disclose use in learning and care in line with policy.
Build and maintain AI literacy. Commit to ongoing, role-appropriate learning, resource training accordingly, and feed experience back to improve local policies and practice.
Responsible AI use should complement, not replace, human judgment. Retain professional accountability and verify AI outputs against clinical and educational judgment.
Think critically before speaking, acting, or uploading with an AI system present. Treat AI as a present third party, anticipate persistence and misinterpretation, and choose words and uploads that protect relationships and trust.
Work within the law. Follow privacy, consent, and intellectual property requirements, and do not delegate legally reserved tasks to AI.
Use and share information ethically in AI-supported environments. Understand how data are collected, stored, and shared; obtain appropriate consent; and safeguard privacy and data sovereignty.
Use AI in ways that actively reduce bias and promote equity. Check outputs for bias, report concerns, support auditing and updates, and adapt use as models and data change.
Build equity into AI foundations. Embed equity in design and governance, ensure accessibility, use diverse data, and co-design with affected communities throughout development and evaluation.
Advance sustainable AI in health systems. Use AI in ways that are proportionate to benefit; mindful of impacts on patient care, the workforce, and the environment; and that strengthen rather than erode people and systems.
Feigerlova takes a step back4. She argues that existing guidance tells educators what students should know about AI, but doesn’t tell us how to actually teach. She suggests embedding verification, critical appraisal and ethical reflection directly into everyday clinical teaching rather than treating AI literacy as a separate module. Recomended tactics include short, structured interventions such as case-based discussions and blended delivery of online and in-person sessions.
I think this is a really refreshing take on how to ‘do’ AI. Competency scales and consensus frameworks are only as good as the classroom practice they translate into, and Feigerlova's contribution is a practical guide to what we can all do next semester. I also like this idea because it eases the pressure on all of us to constantly update our AI teaching to reflect a field that is moving at lightning speed; by anchoring AI usage skills in critical appraisal and reflection we are effectively exiting the AI literacy arms race and trusting students to find their own way.
Naeem NK, Naeem ZF, Anwer A. Ethical engagement with artificial intelligence in faculty research education: evaluation of the New Knowledge Help (NKH) approach. BMC Medical Education. 2026. https://doi.org/10.1186/s12909-026-09405-2
Zhu-ge Y, Yao XT, Mei HX, Yao HX, Chen Q. Construction of a faculty competency model for medical simulation education integrated with GenAI: a multi-method quantitative study based on the perspective of medical students. BMC Medical Education. 2026. https://doi.org/10.1186/s12909-026-09734-2
Sonnenberg LK, Wiljer D, Mamdani M, Do V, Tang B, Haroon B, Jaffer B, Maniate J. Principles for Responsible AI in Health Professions Education, Research, and Care: Health CARE-AI Framework Delphi Consensus Study. JMIR Medical Education. 2026. https://doi.org/10.2196/91626
Feigerlova E. How to Teach Generative Artificial Intelligence in Undergraduate Medical Education. The Clinical Teacher. 2026. https://doi.org/10.1111/tct.70420