MIT Report: AI Can Complete Undergraduate Assignments, Prompting Educational Rethink

AI Capabilities Challenge Traditional Academic Assessment

MIT's ad hoc AI committee, co-chaired by professors Eric Klopfer and Samuel Madden, has issued a stark warning: generative AI can now produce credible solutions to almost any written assignment in the undergraduate curriculum1

1

2

2

. The MIT report identifies that AI in education can handle essays, math and science problems, proofs, and coding assignments with reasonable competence3

3

. MIT President Sally Kornbluth called this moment a "watershed for MIT -- and for all of higher education," emphasizing that addressing these profound challenges to traditional education is not optional2

2

.

The committee's findings go beyond simple concerns about AI cheating scandals. In under three years, AI use in teaching has fundamentally altered campus culture at one of the world's most prestigious institutions1

1

. Attendance at office hours has declined, participation in online discussions has fallen, and anecdotal evidence points to fewer study groups forming in dorms and libraries3

3

. These shifts reflect how students are choosing or feeling pressure to shift toward learning and problem-solving with AI rather than through traditional collaborative methods.

Universities Respond With Alternative Assessment Methods

The challenge of AI can complete undergraduate assignments has prompted institutions to explore alternative assessment methods. The MIT report urges instructors to consider oral exams, portfolios, and in-person conversations tied to work done outside class2

2

. Many professors are switching to hand-written essays and placing stronger focus on in-class discussions, requiring students to keep commonplace notes on their reading and learning3

3

.

The University of Chicago Law School adopted a new "AI strategy" banning phones and laptops in first-year classes3

3

. Princeton University dropped its over century-old Princeton University Honor Code tradition, which allowed students to take exams without supervision, after being shaken by an AI cheating scandal3

3

. The University of Sydney has adopted a two-lane assessment model where one lane uses secure, often in-person work to verify what students can do independently, while the second allows relevant tools, including AI, so students learn how to use them in realistic settings4

4

.

MIT Proposes Comprehensive Educational Overhaul

MIT is considering a rethink of how students learn that extends far beyond classroom rules. The institute's committee said MIT should reconsider what students learn, how faculty assess learning, where AI belongs in that process, and where it should be excluded4

4

. The report calls for every class to state clearly whether students may use AI, must use AI for specific work, or must avoid it entirely4

4

.

MIT's Teaching and Learning Lab has published course design examples using this approach. In one language course, students produce their own translation, compare it with an AI-generated version, and then analyze the machine's choices. A data visualization assignment asks students to compare an AI-assisted attempt with one developed through instructor guidance4

4

. The report emphasizes protecting the value of residential education and human relationships, noting that laboratory sessions, oral defenses, team projects, and in-person problem solving provide evidence of how a person thinks that AI cannot easily certify4

4

.

Research Shows Mixed Learning Outcomes

A March 2026 meta-analysis of 35 experimental studies covering 4,193 participants found a moderately positive overall effect from ChatGPT use on learning outcomes4

4

. However, a separate 2026 review of 67 studies found AI could support critical and creative thinking when instructors built it into structured inquiry, reflection, and evaluation, but found signs of cognitive offloading in loosely structured settings4

4

. Stanford's Accelerator for Learning and ETS said in July, after a convening of more than 100 education, research, and policy leaders, that schools should use portfolios, conversations, performance tasks, formative feedback, and demonstrations of competence to provide richer evidence of learning4

4

.

European Regulation Takes Different Approach

While American institutions debate whether to watch students at all, the EU AI Act has already settled the question and focuses on how closely monitoring should occur1

1

. Systems used for monitoring and detecting prohibited behavior of students during tests are classified as high-risk under Annex III of the EU AI Act, alongside admissions and evaluation of learning outcomes1

1

. Emotion recognition in education has been banned outright since February 2025, and the EU can inspect models and fine providers1

1

. The Digital Omnibus pushed high-risk obligations back to 2 December 2027, buying institutions and their software suppliers another sixteen months1

1

. Watch for how MIT's proposed changes influence academic integrity policies at other institutions and whether regulatory frameworks like the EU AI Act shape future approaches to AI use in teaching globally.