AI is no longer a side issue in higher education. At MIT, a June 2026 report from an expert committee argues that the technology is already altering some of the basic routines that make college learning work: office hours, study groups, classroom discussion, research mentorship, and trust between faculty and students.
The committee’s message is not that universities should reject AI. Its guiding idea is “augmentation not automation”: AI should help people do better work, not replace the thinking, mentoring, and struggle through which students learn.
AI use is moving faster than student preparation
The report describes a campus where students already use AI broadly, while formal preparation has not kept pace. A fall 2025 survey by the MIT student newspaper “The Tech” found that more than two-thirds of students considered AI important for their careers, but only about a quarter felt prepared.
That gap matters because students are not only using AI for convenience. They are entering a world where AI competence may shape future work, research, and learning. The committee therefore wants AI literacy woven into introductory courses right away, rather than treated as an optional add-on for advanced students.
At the same time, AI is changing student behavior in ways that concern faculty. The report says fewer students are showing up to office hours, participation in online discussions is down, and informal study groups in dorms and libraries are thinning out.
Those changes are not just social losses. Office hours and study groups are places where confusion becomes visible, peers test each other’s understanding, and instructors learn what students are actually grasping. When students turn first to a chatbot, the answer may arrive quickly, but the human signals around learning can disappear.
Trust is becoming harder to maintain
The committee also identifies a growing trust problem between faculty and students. Faculty members say policing unauthorized AI use is damaging their relationships with students. Students, meanwhile, fear false accusations and question why faculty may use AI for slides, feedback, or grading while limiting student use.
The report is especially skeptical of AI text detection software. It says these tools are unreliable and often flag work by non-native speakers or neurodivergent students as AI-generated. The committee explicitly advises against using them.
Detection tools also risk pushing campuses into a contest with “AI humanizers,” programs that make AI-generated writing appear human-written. In that environment, assessment can become less about learning and more about surveillance, evasion, and suspicion.
Lockdown browsers do not offer a clean solution either. The report says the current generation of exam software that locks down and monitors computers during tests is buggy and feels like surveillance.
Instead of relying on detection, the committee recommends clearer disclosure. For theses and dissertations, every paper must disclose its AI use, and AI may never be listed as a co-author. The committee also recommends transparency rules for faculty use of AI.
The danger is not just cheating, but shallow learning
The report’s concern goes deeper than academic misconduct. It warns that getting the correct answer from a chatbot can create an “illusion of learning.” Students may feel they understand a subject because they can produce an answer, while the underlying reasoning remains weak.
The committee calls this intellectual surrender: students fall back on AI at the first sign of difficulty. That matters because difficulty is often where learning happens. Struggling through a proof, a reading, a design problem, or a draft can build judgment that a finished answer alone cannot provide.
This issue also reaches MIT’s Undergraduate Research Opportunities Program, or UROP. The report says faculty members are starting to consider AI agents instead of students as research assistants. That could threaten a central part of the MIT experience.
The scale of UROP is large: 93 percent of the class of 2025 participated at least once, and 58 percent of faculty served as mentors. The committee argues that the program exists to train students, not to provide cheap research labor. Replacing students with AI agents would weaken its educational purpose.
Course design should come before AI rules
The committee does not call for one blanket AI policy across all courses. It says each course should begin with its learning goals, then design assessments, and only then decide what AI use to allow.
That approach recognizes that different subjects have different relationships to AI. A poetry seminar and a course on mathematical proofs do not need the same rules. A single institute-wide policy could be too loose in some classes and too restrictive in others.
The report recommends moving toward assessment formats that make learning more visible, including:
- oral exams
- semester portfolios
- in-person discussions
- project-based work
It also says every course should state its AI policy in the syllabus and explain the reasoning behind it. That matters because rules are easier to follow when students understand what the course is trying to protect or develop.
Access and evidence complicate the picture
The report also warns that unequal access to AI tools could widen performance gaps. Some students can afford premium AI access, while others cannot. MIT already gives all members access to various AI models through its Parley platform, with faculty and graduate students receiving $30 per month in free credits. But premium subscriptions from OpenAI, Google, and Anthropic cost several hundred dollars a month, putting them out of reach for many students.
Other research cited alongside the MIT report shows why universities are worried. At Harvard, about 87.5 percent of respondents in 2024 said they used AI, with nearly half using it at least every other day. Around 25 percent said AI led them to visit office hours less, ask instructors for help less, and skip assigned readings.
In the UK, usage among full-time students hit 95 percent by the end of 2025, and students from wealthier households used AI more often. Anthropic found that students offloaded higher-order thinking like analysis and creation to Claude in nearly half of all conversations analyzed.
Several studies point to a possible gap between better homework results and weaker independent learning. A Chinese long-term study with more than 26,000 students found that AI use raised homework grades by 18 percent, but exam scores dropped 20 percent after six months. A UC Berkeley study covering more than 500,000 grades showed that the share of A grades in writing- and coding-heavy courses rose by 13 percentage points since ChatGPT launched, concentrated in courses with a high homework share. At Brown, average scores dropped from 96 percent on a take-home exam to 48.6 percent on the follow-up in-person test.
But the evidence is not one-sided. A two-year study at Vrije Universiteit Amsterdam by legal scholar Thibault Schrepel split students into three groups: no AI, AI without guidance, and AI with training. The no-AI group finished last both years. Even students using AI without guidance did better, despite often accepting AI suggestions without question in class.
Taken together, the MIT report points to a practical conclusion: universities need to teach AI use directly, design assessments around real understanding, and rebuild trust through transparency. The challenge is not simply whether students use AI. It is whether colleges can preserve learning, mentorship, and fairness while AI becomes part of everyday academic work.