UC Berkeley computer science failure rates spike as professors cite AI dependence and weak math skills

Failing rates in introductory computer science courses at the University of California, Berkeley, surged to unprecedented levels in spring 2026, driven by widespread student reliance on artificial intelligence and declining baseline math skills. The trend threatens foundational STEM education at a top-tier public institution, forcing faculty to reconsider admissions standards and classroom policies.

A sharp rise in failing grades

Data from the campus statistics platform Berkeleytime shows that 35.3% of students in COMPSCI 10 and 10.6% in COMPSCI 61A earned failing grades this past spring. In the previous two years, fewer than 10% of students in either introductory course received an F. These figures drastically exceed the electrical engineering and computer sciences department's grading guidelines, which aim to keep D and F grades combined below 7% for lower-division courses. Consequently, average grades in both classes dropped to approximately 2.3 on a 4.0 scale, down from a historical range of 2.8 to 3.3.

AI dependence undermines exam performance

Professor Dan Garcia, who teaches both courses, said in an interview with the student newspaper Daily Californian that the "primary driver" of these rates is a "vast increase in academic dishonesty." Students increasingly used large language models like ChatGPT, Claude, and Google Gemini to complete take-home assignments. This dependency left them without a solid grasp of core concepts, rendering them unable to perform on strictly monitored exams. In CS 10 alone, nearly 30 students were caught cheating on take-home assessments.

For educators managing these shifts, structured guidance on an AI Learning Path for Teachers offers practical strategies for addressing academic integrity. Faculty must now design assessments that require students to demonstrate independent problem-solving rather than relying on automated text generation.

A crisis in mathematical preparedness

Beyond AI reliance, faculty report severe gaps in students' mathematical foundations. Professor Gireeja Ranade said her course assumes prior knowledge of linear algebra, vector calculus, and mathematical proofs. During office hours, she found many students struggled with basic linear algebra. One student admitted a previous UC Berkeley course allowed unrestricted use of the internet and AI tools for both assignments and exams, a policy the mathematics department later denied authorizing.

Faculty push for standardized testing

The mounting academic issues prompted more than 1,400 University of California faculty members to sign an open letter demanding the reinstatement of standardized test requirements for STEM programs. Released on May 25, the petition argues that current admissions practices fail to provide a reliable check on mathematical readiness. According to the letter, "Over the past five years, we have seen a widening divergence in mathematical preparation levels within the same classroom. This trend indicates that current admissions practices do not provide a sufficiently reliable check on mathematical readiness for STEM majors."

Faculty cited a November report from the UC San Diego Senate-Administration Workgroup on Admissions to underscore these concerns. That report found the number of first-year students entering with middle-school-level math skills increased nearly 30-fold since 2020. Roughly one in 12 members of the entering cohort performed below those basic standards. Despite these academic challenges, UC Berkeley remains highly prestigious, holding the No. 1 spot among Top Public Schools in the 2026 U.S. News & World Report rankings.

Why this matters for educators

This situation highlights the immediate need for institutions to audit how AI for Education is integrated into foundational coursework. When students bypass the struggle of learning core concepts, they fail assessments that require independent problem-solving. Educators must establish clear boundaries for AI use in assignments while reinforcing the manual skills required to succeed in proctored environments.