Berkeley CS Courses See Spike in Failing Grades Amid AI Use

The University of California, Berkeley, has recorded a significant and unusual increase in failing grades in several computer science courses during the spring 2026 semester, with instructors citing student overreliance on artificial intelligence, a lack of mathematical preparedness, and understaffing as key factors.

Data from the coursework platform Berkeleytime shows that 35.3 percent of students in the introductory course CS 10, described as a gentle introduction to computer science, received an F. In CS 61A, another lower-division course, the failure rate was 10.6 percent. In the previous two spring semesters, the percentage of Fs for these classes did not exceed 10 percent. The Electrical Engineering and Computer Sciences department’s guidelines state that approximately 7 percent of students in such lower-division courses should receive Ds or Fs. The average grades for both courses in spring 2026 were C-pluses.

Dan Garcia, a teaching professor who taught both CS 10 and CS 61A, identified a vast increase in academic dishonesty linked to student use of large language models like ChatGPT, Claude, and Google Gemini as the primary driver. He stated that nearly 30 students in CS...