MIT Study Finds AI Models Capable of Solving Virtually All Undergraduate Coursework

An ad hoc committee at the Massachusetts Institute of Technology has concluded that modern artificial intelligence tools can generate convincing responses for nearly all written undergraduate assignments. According to findings details first reported by The Next Web, the systems demonstrate competence across a wide array of disciplines, including humanities essays, science and mathematical problems, formal proofs, and programming assignments.

The committee emphasized that the proliferation of AI has altered student behavior well beyond standard concerns over academic integrity. Over the past three years, the rapid adoption of these technologies has driven noticeable structural shifts in campus culture and student collaboration.

Key indicators of academic engagement have seen measurable declines across MIT. The committee recorded lower attendance at faculty office hours, decreased participation on online course discussion platforms, and qualitative reports indicating fewer peer study sessions taking place in residence halls and libraries.

Other top-tier American higher education institutions have initiated policy overhauls in response to similar challenges. The University of Chicago Law School instituted a complete ban on laptops and smartphones in first-year classes, while Princeton University formally retired an honor code system that had guided academic policy for more than a century.

While domestic institutions are increasingly turning toward heightened physical oversight, European authorities are taking a regulatory approach to digital surveillance in academics. Under the European Union’s AI Act, software created to monitor student behavior during exams is classified as high-risk under Annex III, placing proctoring software in the same stringent category as admissions algorithms and grading systems.

The enforcement schedule for these high-risk provisions was recently delayed. Following updates via the EU's Digital Omnibus, compliance requirements for institutions and edtech software vendors were postponed to December 2, 2027, effectively providing a 16-month grace period for compliance.

Certain restrictions under the EU framework remain actively enforced without delay. Outright bans on emotion recognition technology within educational environments took effect in February 2025, granting regulatory bodies the power to audit software architecture and penalize non-compliant developers. The contrast highlights differing global approaches, as U.S. schools struggle with proctoring strategies while European law establishes rigid limits on educational tracking software.