UC Berkeley computer science courses see failure rates surge amid student AI reliance
Failure rates in computer science courses at UC Berkeley surged to over 35% in spring 2026, with faculty attributing the spike to students' heavy reliance on AI language models for coursework. Some students were caught using AI in violation of academic integrity rules, while others became unprepared for exams due to over-dependence on the tools.
What happened
More than 35% of students failed CS 10, an entry-level computer science course at UC Berkeley during the spring 2026 semester, compared to a typical failing rate of 7%, according to an analysis by the Daily Californian.
Teaching professor Dan Garcia told the Californian that he blamed students heavily relying on artificial intelligence models to complete coursework, and said nearly 30 students in CS 10 were caught cheating on take-home exams using AI.
Reported impact
Garcia noted that while some students were caught violating academic integrity rules with AI use, others leaned so heavily on large language models that they were unprepared during exams and struggled to solve problems independently.
Index analysis
Professor Gireeja Ranade noted that one student reported their linear algebra class had an 'open-internet, open-AI policy' for homework and exams, suggesting institutional confusion about AI use guidelines.
Governance controls
The UC Board of Regents eliminated standardized testing requirements for admissions in 2020, which some faculty linked to mathematical unpreparedness among incoming students, compounding issues with AI reliance during instruction.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.