A new report from an MIT ad hoc committee has concluded that artificial intelligence can now credibly complete almost any written assignment in the university's undergraduate curriculum. The findings, published this week, cover essays, mathematics and science problems, proofs, and coding assignments, suggesting that the capabilities of large language models have reached a level where they can produce plausible solutions across a wide range of academic tasks.
The report is not limited to a narrow set of subjects. It explicitly lists written essays, quantitative problem sets, formal proofs, and programming exercises as areas where AI-generated responses are often indistinguishable from student work. This broad scope has significant implications for how universities design assessments, maintain academic integrity, and prepare students for a world in which AI tools are ubiquitous.
Key Findings of the MIT Report
The MIT committee was convened to examine the impact of generative AI on teaching, learning, and assessment. Its central conclusion is that AI models have crossed a threshold: for the majority of undergraduate assignments, these systems can generate credible, contextually appropriate answers. The report does not claim that every AI response is perfect or that all subjects are equally affected, but it does indicate that the default assumption of many professors—that AI cannot handle complex or creative tasks—is no longer tenable.
According to the report, the most immediate concern is not outright cheating. Instead, the committee highlights a subtler but more pervasive shift in campus culture. In less than three years since the widespread availability of generative AI, the technology has altered how students engage with their coursework and with each other. Attendance at office hours has declined, participation in online discussion forums has dropped, and there is anecdotal evidence that study groups in dormitories and libraries have become less common. These changes suggest that students are turning to AI tools for help that they previously sought from professors, teaching assistants, or peers.
The report frames this cultural shift as a challenge to the traditional model of higher education. Office hours are designed not only to answer questions but also to foster mentorship and intellectual community. Online discussions encourage collaborative learning and the articulation of ideas. Study groups build interpersonal skills and deepen understanding through explanation and debate. If AI replaces these interactions, students may miss out on important developmental experiences, even if their grades do not suffer.
Reactions from Other Universities
MIT is not alone in grappling with these issues. Other institutions have already taken concrete steps to respond to the challenges posed by AI. The University of Chicago Law School, for example, has banned phones and laptops in first-year classes, a move intended to force students to engage directly with the material and with each other rather than relying on digital tools. Princeton University has dropped an honor code that had been in place for more than a century, signaling a fundamental reassessment of how academic integrity can be maintained in an era of AI.
These responses reflect a broader tension between trust and verification. The traditional academic model relies on the assumption that submitted work is the student's own. When that assumption is no longer reliable, institutions must choose between trusting students and monitoring them more closely. The direction of travel, as the MIT report and the actions of other universities suggest, is toward supervision. If you cannot trust the work, you watch the person doing it—an old answer to a new question.
The Shift Toward Supervision
The move toward supervision raises practical and ethical questions. Proctoring software, browser lockdowns, and AI-detection tools have become more common in universities, but they are not without controversy. Critics argue that such technologies can be invasive, biased, and ineffective. AI detectors, in particular, have been shown to produce false positives, especially for non-native English speakers. There is also concern that surveillance can create a climate of distrust that undermines the very learning outcomes universities seek to achieve.
Despite these concerns, the trend is clear. Institutions are experimenting with a range of measures, from in-person exams to redesigned assignments that require personal reflection or real-time demonstration of skills. Some professors are incorporating AI into their pedagogy, teaching students how to use it as a tool rather than a substitute for thinking. Others are returning to oral exams and in-class writing to ensure that students can demonstrate their knowledge without external assistance.
European Regulatory Landscape
While American universities debate whether and how to supervise students, European law has already provided a framework. The European Union's Artificial Intelligence Act (AI Act) treats education as a high-risk area, subject to strict requirements. Systems used to monitor and detect prohibited behavior of students during tests are explicitly classified as high-risk under Annex III of the legislation. This classification carries obligations related to transparency, data governance, and human oversight, among others.
The AI Act's obligations were originally due to apply in a certain timeframe, but the European Commission's Digital Omnibus package has pushed the deadline to 2 December 2027, giving educational institutions and software suppliers an additional sixteen months to comply. However, the prohibitions in the AI Act were not delayed. Emotion recognition in educational settings has been banned outright since February 2025, and the EU has the power to inspect AI models and fine providers that violate the rules.
This regulatory environment means that European universities must be particularly careful when deploying AI-based monitoring tools. The use of such tools is not prohibited per se, but it is strictly regulated. For example, a system that flags suspicious behavior during an online exam would need to meet high standards of accuracy, fairness, and data protection. The ban on emotion recognition goes further, prohibiting any system that attempts to infer a student's emotional state from facial expressions or other biometric data.
AI and Academic Integrity: A Broader Context
The debate over AI in education is part of a larger conversation about the role of technology in learning. Generative AI models, such as those based on large language models, have become increasingly capable since the public release of tools like ChatGPT in late 2022. These tools can generate text, code, and even mathematical proofs that are difficult to distinguish from human output. As the technology continues to improve, the boundary between legitimate assistance and academic dishonesty becomes harder to draw.
Some educators argue that AI should be embraced as a learning aid, much like calculators or spell-checkers were in earlier decades. They point out that students need to learn how to work with AI, because it will be part of their professional lives. Others maintain that the ability to reason from first principles is a core outcome of a university education, and that outsourcing that thinking to AI undermines the entire enterprise.
The MIT report adds empirical weight to these debates by demonstrating that AI is no longer a hypothetical challenge but a present reality. The fact that AI can complete almost any undergraduate assignment raises questions about the validity of many traditional assessment methods. If a machine can produce a passing essay or a correct proof, what exactly does that assignment measure?
Implications for Students and Educators
For students, the availability of AI tools creates both opportunities and risks. On one hand, AI can serve as a personalized tutor, explaining difficult concepts, providing feedback on drafts, and helping with debugging. On the other hand, the temptation to use AI without genuine engagement can lead to shallow learning and a false sense of accomplishment. The MIT report's observation that office hours and study groups are in decline suggests that many students are choosing the path of least resistance, even if it means sacrificing deeper learning.
For educators, the challenge is to design assessments that are resistant to AI while still fostering critical thinking and creativity. Some universities are experimenting with "flipped" classrooms, where students watch lectures at home and use class time for discussion and problem-solving. Others are incorporating AI into assignments in ways that require students to critique or improve AI-generated outputs, thereby building skills that are directly relevant to modern workplaces.
The regulatory landscape in Europe adds another layer of complexity. Universities must navigate the AI Act's requirements while also ensuring that they do not inadvertently violate students' rights to privacy and non-discrimination. The emphasis on human oversight means that automated systems cannot make final decisions about student assessment or behavior without human involvement. This is a significant departure from the earlier enthusiasm for AI-driven proctoring, which often operated without meaningful human validation.
The Future of Assessment
Looking ahead, the MIT report and the European regulatory framework suggest a future in which AI will play a more visible role in higher education, but not necessarily in the way many imagine. It is unlikely that universities will simply ban AI or return to pen-and-paper exams in all subjects. Instead, they will need to develop new forms of assessment that are both authentic and resistant to abuse. This might involve more oral examinations, project-based learning, and collaborative assignments that cannot be easily replicated by a machine.
It is also likely that universities will become more sophisticated in their use of AI for detection and support. The EU's AI Act does not prohibit monitoring technology per se, but it demands rigorous testing and accountability. This could lead to the development of more reliable and fair tools, perhaps even ones that outperform human proctors in detecting certain kinds of misconduct.
At the same time, the cultural shifts identified by MIT may be irreversible. The Generation Z and Generation Alpha students who are now entering universities have grown up with AI. They are accustomed to asking a chatbot for help with homework, drafting emails, or generating ideas. Universities that fail to adapt to this reality will struggle to remain relevant. Those that embrace it, while setting clear boundaries and ethical guidelines, may find that AI becomes a valuable ally in the mission of education.
For now, the debate continues. Academic institutions across the world are watching MIT's findings closely as they develop their own policies. The report's conclusion—that AI can complete almost any undergraduate assignment—is both a warning and an invitation. It warns that old methods of assessment are becoming obsolete. It invites educators to think creatively about how to Foster genuine learning in an age of intelligent machines.