Ask The PhD Community

HigherEd AI Daily

September 21 – Governing AI on Campus: From Sacramento to Redmond

Monday, September 21, 2026

Today's edition examines AI governance moving from principle to enforceable policy, what a sweeping global survey reveals about student workforce anxiety, and a principled framework for understanding responsible AI writing in academic life.

The Neuron Daily — GOVERNANCE

California Governor Orders Stronger Frontier-AI Safety Rules Within Two Months

California Governor Gavin Newsom has signed an executive order directing policy experts to recommend stronger regulatory frameworks for frontier artificial intelligence systems, with recommendations due within two months. The order specifically calls for proposals that could include independent safety review plans and emergency kill-switch provisions for the most powerful AI models. The directive follows a series of high-profile incidents in which AI systems operated outside intended parameters and signals that state-level governance is accelerating ahead of federal action.

California has long served as the de facto regulatory proving ground for technology policy in the United States. The state's track record with legislation such as the California Consumer Privacy Act means frameworks that emerge from this executive process are likely to influence national and international conversations. The compressed two-month recommendation window suggests that formal rule-making proposals could follow as early as the first quarter of 2027; any formal rule-making process would typically include a period for public and institutional comment before regulations are finalized.

Higher education institutions, particularly those operating large AI research centers or deploying commercial frontier models at scale, have a legitimate stake in how the state defines the scope of these requirements. Engaging the process early gives campus voices a seat at the table while the definitions are still being written.

Why it matters for campuses

A California mandate for independent AI safety plans and emergency shutdown provisions would directly affect institutions operating large-scale AI infrastructure or using frontier models through commercial providers. Chief information officers, research computing directors, and university counsel at California campuses or institutions with California-based research centers should begin cataloguing which AI systems in their environment might fall under any forthcoming definitions of a regulated frontier model. Submitting institutional comment during any formal rule-making period is one of the clearest channels higher education has to shape a regulatory outcome before it is fixed.

Read More

The Neuron Daily — POLICY

Microsoft Opens Six-Week Public Comment Window on Draft AI Code of Conduct

Microsoft has opened a six-week public comment period on its draft AI code of conduct, accepting submissions through late October. The document outlines governing principles for how the company will develop and deploy AI systems, with specific provisions covering human control over AI operations, scoped permissions for autonomous agents, mandatory shutdown protocols, and bounded operational goals. The public comment structure is unusual for a private technology company and reflects growing pressure from governments and civil society to make AI governance decisions more transparent and participatory.

The code is directly tied to Microsoft's forthcoming "MAI" model family, positioned as enterprise-grade AI systems. The draft language centers on containment; each AI system is expected to operate within defined limits, request explicit permissions before taking consequential actions, and support the ability of authorized personnel to pause or halt operations at any time. The terminology around "human control" and "bounded goals" mirrors language increasingly common in AI governance frameworks being developed by universities, professional associations, and regulatory bodies worldwide, making this comment window an opportunity to align institutional positions with a commercial framework that will affect campuses directly.

Why it matters for campuses

Institutions that rely on Microsoft 365, Azure AI services, or Copilot-integrated products have a direct interest in how the company defines human oversight and permission boundaries in AI contexts. Faculty using Copilot in Word, administrators relying on AI-assisted scheduling, and researchers using Azure cognitive services will all operate under whatever framework emerges from this process. University AI ethics committees, faculty governance bodies, and research computing offices have standing to submit comment; doing so as an institution or scholarly organization is one of the few direct channels higher education has to shape the conduct of a major AI provider during its policy formation period. The window closes in late October.

Read More

TLDR Marketing — RESEARCH

Pew Research: In 34 of 37 Countries, More People Expect AI to Eliminate Jobs Than to Create Them

A new Pew Research Center study spanning 37 countries found that in 34 of them, more people expect artificial intelligence to cause job losses than to create new employment opportunities. The survey is among the most geographically extensive cross-national studies of public AI attitudes conducted to date; it reveals that labor displacement anxiety is not confined to any single region, economic context, or political tradition. Majorities or pluralities across North America, Europe, Asia, and Latin America reported expecting net negative employment effects from AI in the coming years.

The findings diverge from some economic projections that emphasize AI's job-creation potential in emerging technical fields. Pew researchers note that public expectations and measured outcomes can differ; historical patterns suggest technology-related employment concerns sometimes run ahead of observable disruption. Nonetheless, the breadth and consistency of negative expectations across 34 of 37 surveyed countries indicates that this is a durable feature of the current moment rather than a localized or passing reaction. Workforce expectations in the survey closely mirror what campus career centers, professional schools, and workforce development programs report hearing from students week after week.

Why it matters for campuses

Career services offices, academic advising departments, and workforce education programs face a student population that is actively anxious about AI-related job displacement, and this Pew data provides credible empirical grounding for those conversations. Colleges revising or developing workforce programs in fields with high automation exposure, including paralegal work, entry-level writing, data entry, and customer service, should treat this survey as a signal that students need explicit preparation for navigating AI-transformed labor markets rather than reassurance that the concern is overblown. Program designers and institutional researchers will find the country-level data useful for making the case for curriculum investment to senior leadership.

Read More

TLDR Dev — TOOLS

A Principled Critique of AI-Generated Documents: Why Authorship Still Demands Intellectual Presence

A widely circulated essay by software engineering leader Colin Breck argues that AI-generated professional documents often shift cognitive responsibility rather than genuinely sharing authorship. Documents produced through heavy AI involvement tend to bury purpose, risk, and the author's actual judgment beneath fluent but hollow prose; readers are left sorting through text that only the prompter had enough context to evaluate, while the person who produced it never fully wrestled with the substance. The concern is not grammatical quality or factual accuracy in isolation but the absence of an identifiable mind behind the argument.

Breck proposes a workable alternative: use AI to verify facts, surface omissions, complete citations, and tighten awkward prose, while holding the human author responsible for every sentence. The goal is not to restrict AI use but to preserve the epistemic accountability that makes a document worth reading and, in professional or academic contexts, worth trusting. Writing that no one fully stands behind is noise, even when it is impeccably formatted.

The practices Breck critiques, prompting AI for a first draft and lightly editing the result, are already common among some faculty and widespread among students. In academic settings, the consequence is direct: if the person who submits a document has not genuinely engaged with its claims, peer reviewers, instructors, and dissertation committees cannot trust that the document represents anyone's actual understanding of the subject.

Why it matters for campuses

Academic integrity policies have largely focused on detection rather than principle, which tends to produce an arms race between students and software rather than genuine clarity about what honest AI use means. The framing in this essay offers a more durable anchor: the question is not whether AI was used but whether the human author can be held intellectually responsible for every claim in the work. Faculty designing assignments, writing instructors revising rubrics, and graduate programs developing AI-use policies will find this standard more resilient and harder to game than detection-based approaches alone; it also gives students a clear affirmative standard to aim for rather than a list of prohibited tools.

Read More

Tool of the Day

AgentCloak

AgentCloak sits between your organization and any AI model, automatically replacing sensitive details with safe stand-in values before a prompt leaves your network and restoring the originals only for authorized users after the model responds. The tool works across ChatGPT, Claude, Gemini, and other major AI assistants and is designed for organizations that need to use AI at scale while keeping protected information, including names, identification numbers, and confidential records, out of third-party model training logs and data pipelines. For higher education specifically, the most immediate application is supporting FERPA-conscious AI use across faculty and administrative workflows.

Try it: Route a midterm feedback draft through AgentCloak before sending it to Claude or ChatGPT, allowing the tool to substitute placeholder tokens for actual student names and ID numbers; review the model's suggestions, then have AgentCloak restore the personalized version internally so the feedback is ready to send without any protected information ever reaching the model provider.

Visit AgentCloak

Have a great learning day!

Dr. Ali Green

Sources for This Edition

The Neuron Daily (theneurondaily.com)
TLDR Marketing (tldrnewsletter.com)
TLDR Dev (tldrnewsletter.com)
CNBC (cnbc.com)
Decrypt (decrypt.co)
Pew Research Center (pewresearch.org)
Colin Breck's Blog (blog.colinbreck.com)

askthephd.com
 | 
askthephd.substack.com
 | 
Unsubscribe

HigherEd AI Daily; Curated by Dr. Ali Green