|
Ask The PhD Community
HigherEd AI Daily
May 26 – Governance, Guardrails, and a Curriculum in Flux
|
|
Tuesday, May 26, 2026
Higher education's most pressing AI questions converged this week around governance, policy risk, curriculum disruption, and the future of institutional knowledge.
|
|
The Rundown AI — GOVERNANCE
Pope Leo XIV Issues 42,000-Word Manifesto on AI Governance
Pope Leo XIV released "Magnifica Humanitas," a 42,000-word encyclical addressed to the Catholic Church's 1.4 billion members, drawing a direct parallel between the challenges posed by artificial intelligence and those of the Industrial Revolution. The document argues that AI is "never neutral" and that its primary drivers are private, transnational companies whose reach already surpasses the regulatory capacity of many national governments. The Pope called for robust legal frameworks and independent oversight, and stated plainly that lethal decisions must never be delegated to algorithms: "no algorithm can make war morally acceptable."
The encyclical's reach extends well beyond religious institutions. By naming the structural incentives of frontier AI development as a core ethical concern, the document enters the same territory that academic ethicists, governance scholars, and campus policy committees have been navigating for years. Notably, Anthropic researcher Christopher Olah publicly aligned with the Pope's framing, acknowledging that every frontier AI lab operates inside incentives that can conflict with responsible development.
The document represents the most comprehensive institutional statement on AI ethics issued by any major global institution to date; its framework of independent oversight, legal accountability, and explicit rejection of autonomous lethal decision-making is directly relevant to policy conversations happening on campuses right now.
Why it matters for campuses
University leaders drafting or revising AI governance policies now have an unusual and significant external reference point. The encyclical's framework maps directly onto the principles many institutions have been struggling to articulate; faculty in theology, philosophy, law, and public policy have immediate course material, while provosts and AI task forces have a new touchstone for conversations with boards, accreditors, and government funders.
Read More
|
|
The Rundown AI — POLICY
Open-Source AI Safety Guardrails Can Be Stripped in Under Ten Minutes, FT Investigation Finds
A Financial Times investigation revealed that a publicly available GitHub tool called Heretic removes safety guardrails from open-source AI models using just four lines of code and no specialized hardware. FT reporters stripped Meta's Llama 3.3 in under ten minutes; the tool's creator stripped Google's Gemma 4 within 90 minutes of its public release. Heretic has since produced more than 3,500 modified models that have been downloaded over 13 million times.
The implications are immediate and troubling. Modified models in the investigation responded to questions about bioweapons and child exploitation without restriction. Google described the situation as "a known technical challenge facing all open models"; Meta declined to comment. The story underscores a deepening gap between the safety assurances AI companies make at model release and the practical reality of open distribution.
Why it matters for campuses
Campuses that have adopted or are evaluating open-source AI tools for research, instruction, or student use face a direct governance challenge. The ease with which guardrails can be removed means that a model approved for institutional deployment today may not behave as expected if students or researchers obtain modified versions elsewhere; academic IT offices, research computing teams, and AI ethics committees should treat this investigation as a concrete case study for their institutional risk frameworks.
Read More (subscription may be required)
|
|
TLDR Dev — TEACHING
The Death of the Programming Textbook and What It Means for CS Education
Sales of physical programming books are collapsing. A widely circulated essay argues that generative AI tools and automated coding assistants have effectively replaced printed technical manuals, offering instant, context-specific answers that a static text cannot match. The author does not celebrate the shift; the piece frames it as a fundamental change in how developers acquire knowledge, noting that books once fostered deep, systematic understanding in ways that AI-assisted lookup does not replicate.
The trend is not limited to casual readers. Publishers of foundational computer science texts report sustained declines across categories ranging from introductory programming to advanced systems design. Learners now expect an interactive, responsive tool rather than a structured curriculum delivered through sequential chapters; the essay argues this shift represents a move to higher abstraction levels and away from first-principles learning.
Why it matters for campuses
Computer science and information technology programs are watching a key pedagogical assumption dissolve in real time; if students no longer engage with structured technical literature, faculty face urgent questions about how to teach foundational knowledge and deep problem-solving in an environment where AI provides an answer before the student has worked through the reasoning. Curriculum designers and department chairs should engage this trend directly rather than wait for enrollment or assessment data to surface the problem.
Read More
|
|
TLDR Dev — RESEARCH
Norway's National Library Is Training a Sovereign LLM to Preserve Its Language and Culture
Norway's National Library is training a large language model on approximately two petabytes of digitized national records, explicitly designed to preserve and extend understanding of Norway's language, history, and culture. The project is built and governed domestically rather than through external commercial providers, representing one of the clearest examples to date of a public institution treating AI infrastructure as a matter of cultural sovereignty rather than pure technical utility.
Rather than relying on a general-purpose commercial model trained predominantly on English-language data, Norway is investing in a system that reflects its specific linguistic and historical context. The project's scale and institutional framing set it apart from similar national AI efforts that have remained at the proof-of-concept stage.
Why it matters for campuses
The Norway project is a direct model for what higher education institutions could pursue at scale; academic libraries hold vast, unique collections, including digitized primary sources, oral histories, and specialized corpora that would never appear in a commercial training dataset. Institutions thinking about long-term research infrastructure should examine whether building or contributing to domain-specific or regionally sovereign AI systems is a strategic priority rather than an afterthought.
Read More
|
|
Tool of the Day
Synthesia
Synthesia converts written text, documents, and scripts into narrated video presentations using AI avatars, typically in ten to fifteen minutes. Educators can upload a PDF, paste course content, or provide a structured prompt and receive a video with narration, scene transitions, and customizable visual styles. The platform is widely used in corporate training, but its core capabilities map directly onto instructional design, continuing education, and research communication tasks.
Try it: Upload a module overview or reading summary from one of your current courses and prompt Synthesia to generate a two-minute introductory video for students; use the audience and objective fields to specify undergraduate learners and conceptual introduction as your goals.
Visit Synthesia
|
|
Time to rethink how you are delivering your textbook. Old school? I still love that. Notetaking. I still love that too! We are learn in different ways. However, for some, videos, simulations, and interactions help students learn the content.
Dr. Ali Green
|
|
Sources for This Edition
The Rundown AI (daily.therundown.ai)
TLDR Dev (tldrnewsletter.com)
|
|
askthephd.com
|
askthephd.substack.com
|
Unsubscribe
HigherEd AI Daily; Curated by Dr. Ali Green
|
|