Years from now, we may look back and ask what the artificial intelligence (AI) ecosystem looked like at this moment. This essay offers one such view. It is not exhaustive, nor is it an attempt to predict outcomes. Instead, it reflects on how AI is absorbed into existing institutions—and how that process is shaped as much by power, incentives, and uneven capacity as by technical capability.

AI’s role in higher education cannot be understood in isolation. Universities encounter AI within a broader ecosystem that includes technology firms, federal agencies, national laboratories, and national security priorities. These relationships are uneven by design. Institutions differ widely in their access to advanced tools, reliance on external platforms, exposure to legal and ethical risks, and capacity to shape the direction of AI research rather than merely adapt to it. Some universities help set the agenda; others inherit it.

For that reason, higher education offers a useful lens through which to examine the broader AI moment. While AI’s uneven effects extend well beyond the academy—reshaping labor markets, generating intellectual property disputes, and producing stark disparities in access and deployment across countries and communities—there are data specific to colleges and universities that help illustrate both the speed of adoption and the inequalities that accompany it. What follows traces several of these developments, not to catalogue every educational use of AI, but to highlight the patterns that emerge when powerful technology is rapidly absorbed into an already stratified system.

A Cengage report noted AI adoption rates of 63 percent among K-12 teachers and 49 percent among higher education instructors. The adoption rates are likely higher, as the year-old data may conceal the rapid spread of AI. A more detailed look shows that higher education instructors use generative AI:  

[T]o create course content / student-facing materials (45%; +11% from 2023) including quizzes and assessments (39%; +16% from 2023), assist in lesson planning (42%; +18% from 2023) and support their lectures (42%; +12% from 2023). Nearly 2 in 5 (36%) use it to complete administrative tasks (36%; +3% YoY). HED students are also leveraging GenAI in learning, primarily by using it to help summarize complicated concepts (67%), generate writing assignment ideas (61%) and create study materials (55%).  

Microsoft, through its work on Copilot, observed that the benefits of AI use varied across users and differed significantly between performance on tests and on assignments. Notably, these differences were not linked to socio-economic status.

Higher education does not exist in isolation from AI’s technological effects. Governing Boards, such as the Carsey-Wolf Center at UC Santa Barbara, recognize this revolution and envision how their institutions should respond. At a recent panel on the media industry and AI, Rick Rosen, head of TV at WME, spoke about the direction the Carsey-Wolf Center should take. See the clip below:

Higher education is also affected by federal funding. We can see a drift in emphasis from science funding to the Department of Energy’s new Genesis Mission. Science now has a techno-optimistic lens. The latest AI Executive Order conceives of this national effort as a Manhattan Project rather than a moon landing; we hear notes of a military objective in this mission, which has shifted from nuclear development to AI. Dario Gil, the Genesis Mission Director, ties this federal effort to related work in universities and private laboratories, saying:

The Genesis Mission marks a defining moment for the next era of American science. We are linking the nation’s most advanced facilities, data, and computing into one closed-loop system to create a scientific instrument for the ages, an engine for discovery that doubles R&D productivity and solves challenges once thought impossible.

Equally important is how we, as instructors, convey understanding of AI to students. Jensen Huang, CEO of NVIDIA, reminds us of how AI “thinks.” I imagine instructors including his remarks in class discussions about AI. Huang delivered his comments at Cambridge Union upon receiving the 2025 Hawking Fellow Award. See the clip below:

Over the course of 2025, I encountered annoying observations nibbling around the edges of these grand visions.

For example, AI needs vast amounts of data. There are outstanding conflicts surrounding data scraped from the internet and various datasets. The issue of AI fair use raises questions about whether all this data can be used freely or requires compensation. Numerous cases are in litigation over this issue. Some companies have already settled. Anthropic settled a complaint involving 500,000 books for $3,000 per book. Had the case gone to trial, the resulting damages could have been far greater, with more significant consequences for the corporation.

Additionally, the emergence of an “evil” agent may be even more worrisome.

An Anthropic alignment research group published Natural Emergent Misalignment From Reward Hacking in Production RL. The term reward hacking, which dates to the early 1980s, refers to the tendency of reinforcement-learning systems to develop shortcuts—or cheat—that technically satisfy their reward function while undermining its intended goal.The Anthropic example with Claude found that “safety training using standard chat-like prompts results in aligned behavior on chat-like evaluations, but misalignment persists on agentic tasks.” The misalignment also includes sabotage. The researchers described this problem more dramatically in a podcast. See the clip below:

And so this review arrives at a conclusion many others have reached in one form or another: we are very good at running fast—including in our adoption of AI in education—but far less certain about where we are running.

The forthcoming second installment of this two-part essay advances an AI paradigm for addressing complex social science challenges. As AI models continue to mature, the academy may come to see AI collaborators not as optional tools, but as essential components of scholarly inquiry.