UC Berkeley's computer science program is experiencing an unprecedented academic crisis, with failing grades in introductory courses reaching 35.3% in spring 2026—more than triple historical norms. The electrical engineering and computer sciences (EECS) department attributes the spike primarily to widespread academic dishonesty driven by large language models like ChatGPT, Claude, and Google Gemini.
Spring 2026 Failing Rates Exceed Department Guidelines by 400%
According to Berkeleytime statistics, CS 10 saw 35.3% of students receive F grades in spring 2026, while CS 61A recorded a 10.6% failure rate. These numbers represent a dramatic increase from spring 2025 and spring 2024, when neither course exceeded 10% failing grades. The EECS department's official grading guidelines recommend just 7% of students in lower division courses receive D's and F's combined.
Upper division courses showed similar patterns. EECS 127 recorded a 16.8% failure rate—more than three times the department's typical 5% benchmark for D's and F's in advanced courses.
Professors Cite AI-Driven Academic Dishonesty as Primary Driver
Dan Garcia, an EECS professor at UC Berkeley, identified the "primary driver" of abnormally high failing rates as a "vast increase in academic dishonesty" stemming from student reliance on large language models. Students are increasingly using AI tools to complete assignments and exams, undermining their ability to develop fundamental programming and problem-solving skills.
Professors Gireeja Ranade and Jelani Nelson are also involved in addressing the crisis. Beyond AI misuse, instructors point to declining mathematical preparedness among incoming students and department understaffing as contributing factors to the grade deterioration.
Implications for Elite Computer Science Education
The crisis at UC Berkeley—one of the world's top computer science programs—raises questions about how AI tools are reshaping undergraduate education. The spring 2026 statistics suggest that traditional assessment methods may be failing to distinguish between student knowledge and AI-generated work, while students who rely on AI assistance may be entering courses unprepared for examinations where such tools are restricted.
Key Takeaways
- UC Berkeley's CS 10 course saw 35.3% failing grades in spring 2026, compared to under 10% in previous years
- The EECS department recommends just 7% D's and F's for lower division courses, making actual rates 5x higher than guidelines
- Professors attribute the crisis primarily to academic dishonesty driven by ChatGPT, Claude, and Google Gemini usage
- Upper division course EECS 127 recorded 16.8% failures, more than three times the typical 5% benchmark
- Declining math skills and department understaffing are cited as additional contributing factors