On October 28, Dartmouth announced that its provost, Santiago Schnell, was leaving the university after a faculty committee examined his undisclosed use of AI in published writing. The same day, University of Minnesota learning scientists warned that casual use of general-purpose AI can threaten the learning processes education depends on. Between October 25 and 31, the question running through the news was the same at every level, from students to senior leaders: which work can be handed to AI, which must stay with people, and who answers for the result? Universities are expanding access to more capable systems for research, teaching, and administration. The harder task is setting clear boundaries for that delegation and holding everyone to them.
“This is not a technology project.”
Kavita Bala, Provost, Cornell University, on the university’s new AI access pilot (Hayes, 2026)
Minnesota learning scientists warn that careless AI use can undermine learning itself
The Details
A University of Minnesota report, led by educational psychology professor Panayiota Kendeou and based on a survey of experts in learning and AI, concludes that generative AI could seriously disrupt key learning processes. The authors warn that “casual, careless use of general-purpose GenAI can threaten the very learning processes” education depends on, with risks to thinking, learning skills, collaboration, social-emotional development, and human connection. They also note that high-quality research on AI’s effects on learning is still scarce. The report urges the most caution in elementary school, where foundational skills are forming, and recommends that students first work independently to build their own ideas and skills, then use AI to test ideas, extend arguments, and find resources once those skills are in place. “With AI in education, we cannot afford to get it wrong,” Kendeou said (Shockman, 2026).
Why it Matters
The report centers on younger learners, but the principle applies directly to college: students should be able to do the work themselves before AI does it with them. A correct AI-assisted answer is not evidence of understanding. Faculty can sequence assignments so independent work comes first, then require students to explain, revise, or defend what AI helped them produce. Programs should identify the assessments where AI-supported performance could hide gaps in what a student actually knows.
Five universities join Project ProtoStar to make public research usable by AI
The Details
George Washington University, Carnegie Mellon University, Purdue University, Georgia Tech, Arizona State University, and the Astera Institute are founding partners in Project ProtoStar, a national effort to make publicly funded science easier for people and AI systems to find, understand, and reuse. Hugging Face is the infrastructure partner and the Open Research Community Accelerator is the community partner. GW’s Open Source Program Office will set up a Hugging Face presence for the university’s research data, code, and models, help researchers use it, and give students in research groups hands-on experience with open models. “Public access is a good first step. Usability is next,” said Lorena A. Barba, faculty director of the office and professor of mechanical and aerospace engineering (George Washington University, 2026). Carnegie Mellon is hosting the first ProtoStar challenge, Climate Risk Intelligence, from November 2 to 13, in which teams use historical temperature and precipitation records from 210 U.S. cities to analyze extreme weather and recommend adaptation strategies (Carnegie Mellon University, 2026).
Why it Matters
Much publicly funded research is technically open but practically hard to reuse because data and code are scattered and poorly documented. ProtoStar works on that infrastructure problem rather than adding another AI tool. For research universities, it raises practical questions about documentation standards, licensing, intellectual property, sensitive data, and whether staff have the capacity to prepare research for responsible reuse. The initiative is new, and its value will depend on whether shared resources lead to credible, reproducible work.
Policy & Governance
-
Bath researchers map the governance risks of AI in university administration
In a conceptual analysis published October 29, Namita Pandey and Dan Davies of the University of Bath School of Management examine AI in front-office university services: recruitment, admissions, enrollment, student records, advising, housing, welfare, and careers. They identify five drivers of adoption: infrastructure readiness, competition under enrollment volatility, regulatory and mobility mandates, commercial platform ecosystems, and sustainability legitimacy. They link administrative AI to visa regimes, data sovereignty, and universities’ financial dependence on international students. Procurement terms, routes to appeal automated decisions, student and staff participation, audits, and public registers of systems, they argue, determine whether these tools strengthen institutions or create new dependencies. The paper draws on documented failures, including the UK’s 2020 Ofqual grading algorithm and a Home Office visa-streaming tool (Pandey & Davies, 2026). It is a framework, not an outcome study. Before using AI in admissions, advising, or student services, decide which decisions stay with people, how errors get corrected, and how students can challenge an outcome.
-
UT Austin requires an AI policy in every syllabus and does not endorse AI detectors
Starting with fall 2026 syllabi, every course at the University of Texas at Austin must include a Course AI Policy Statement designating generative AI as permitted, partially permitted, or prohibited. Faculty decide which applies, and if AI is allowed in any form they must explain what is permitted and how students should disclose their use (University of Texas at Austin, n.d.-b). The university does not endorse AI-detection software and holds no central contract for it. Its UT Sage platform lets instructors build custom AI tutors that students use in Canvas (University of Texas at Austin, n.d.-a). The approach keeps faculty in charge of their courses while guaranteeing students clear expectations before they submit work. Check whether your institution’s course-level AI policies are clear, consistent with program requirements, and backed by examples.
-
Leading Indian institutions move from AI bans to AI by design
As reported by The Indian Express on October 27, several Indian institutions are rewriting rules for AI in teaching and exams. At IISER Pune, generative AI is now allowed by default unless an instructor or doctoral supervisor rules it out in writing; in one bioinformatics course, students could use AI to write data-analysis code but had to solve similar problems unaided on the exam. IIM Ahmedabad bought 1,500 AI-tool licenses and lets students use AI to summarize cases so class time can focus on applying them. At IIM Bangalore, 78 MBA students in the June–July term built chatbots, restricted to course material, that answered their generative AI midterm, and were graded on prompt engineering, parameter tuning, and how well they applied course concepts. ISB leaves the decision to each faculty member and now weights problem framing and decision-making more heavily, backed by in-person case defenses (Elets Digital Learning, 2026). For business schools especially, the line that matters is between using AI as a professional tool and showing independent judgment. Decide which tasks students must do on their own.
Programs, Research & Infrastructure
-
Cornell opens multiple AI tools to its community, with orientation required
Cornell announced October 26 a two-year pilot giving base-level accounts for OpenAI’s ChatGPT and Codex, Anthropic’s Claude Cowork and Claude Code, Microsoft Copilot Chat, and, tentatively, Google Gemini Chat. Faculty, staff, and doctoral students in Ithaca and at Cornell Tech get access starting in November, with master’s and undergraduate students following in early 2027. Faculty and staff must complete an orientation on responsible use, and access to agentic tools comes with a separate primer. The contracts bar using entered data to train AI models. A research program called CoPAIR will develop discipline-specific norms, and a faculty, staff, and student working group will track the tools’ energy use (Hayes, 2026). Cornell is moving AI use out of personal accounts and into managed, trained, and evaluated access, with a built-in point to reassess after two years. That sequence, access paired with training, data rules, and review, is a useful model.
-
AI literacy in creative programs links to well-being as much as performance
Zuji Liang, Yuanxiong Liu, Churan Wang, and Xiaolin Zhang surveyed 475 design and creative students in Asia, Europe, and Australia, in a study published October 30. AI knowledge was strongly associated with awareness, operational ability, and ethical awareness. Awareness and operational skill were linked to critical reflection, which was linked to students’ perceived learning performance. Ethical awareness predicted academic well-being but not perceived performance. The study is cross-sectional and self-reported, measured performance through students’ own perceptions, and several measures were highly correlated, so it does not show that AI literacy training improves outcomes (Liang et al., 2026). In creative fields, ethics guidance on authorship, fairness, and acceptable use should be part of the discipline’s AI preparation rather than a separate compliance exercise.
Other
-
Cornell launches a humanities and social sciences initiative on AI
Cornell’s College of Arts and Sciences announced October 29 a Humanities, Social Sciences + AI initiative. It will fund five postdoctoral fellows for up to two years to study AI’s social, political, ethical, cultural, and economic effects; bring practitioners from industry, government, civil society, and the arts to campus; develop a university-wide AI and Society minor; and host an annual symposium. The first year is funded through the Cornell AI Initiative and a private gift (Glaser, 2026). Universities that build AI capacity only in technical fields leave students unprepared for the judgment questions AI raises in every profession.
Do It Now Checklist
Betting On: Responsible AI Delegation
This week’s stories turn on one question: what can be handed to AI, and who stays accountable? Dartmouth showed that disclosure standards apply to leaders as much as to students. Minnesota’s learning scientists argue students must build skills before handing work to AI. Bath’s framework asks who answers for automated administrative decisions, and Cornell paired broad access with training and review. Access to capable systems is easy to provide. The institutions that do this well will define which responsibilities can be delegated, which cannot, and how they will know the difference.
With Inspiration Moments, we share motivational nuggets to empower you to make meaningful choices for a more fulfilling future. Remember: “Life happens for you, not to you, to live your purpose.”
Respectfully,
Dr. Lynn Austin