Berkeley CS Courses Are Seeing Failing Grades Soar With Greater AI Usage

UC Berkeley computer science courses saw failing grades soar and class averages drop to C+ levels in spring 2026, as professors enforced stricter academic integrity policies and in-person exams amid widespread student use of generative AI tools. The grade collapse, which pushed flagship CS and electrical engineering courses well below the university's expected B-range average of 2.8-3.3 GPA, exposed a gap between AI-assisted homework performance and students' actual mathematical reasoning skills. Researcher Igor Chirikov warned that AI-driven grade inflation undermines the credibility of academic transcripts, making Berkeley's correction a potential preview of higher education's broader reckoning with artificial intelligence.

Generative AI has been a GPA fairy godmother for most college students. Research from UC Berkeley’s own education https://www.universityworldnews.com/post.php?story=20260514074518988 researchers shows AI-exposed courses nationwide saw A grades jump 13 percentage points after ChatGPT https://www.gadgetreview.com/man-uses-chatgpt-to-design-cancer-vaccine-that-saved-his-dogs-life launched, with failure rates dropping from 3% to 2% . Students submit polished homework, professors see impressive work, everyone wins. Except at UC Berkeley’s computer science department https://eecs.berkeley.edu/ , where spring 2026 brought a harsh reality check. Major CS and electrical engineering courses—the gateway classes that determine who stays in tech—posted C+ averages and failure rates that would make pre-med organic chemistry blush. When Standards Meet Silicon Valley Dreams Berkeley’s engineering policies expect B-range averages, not the 2.3 GPAs now appearing. UC Berkeley’s College of Engineering https://engineering.berkeley.edu/students/undergraduate-guide/policies-procedures/grades/ doesn’t mess around with grade distribution guidelines. Lower-division technical courses should average in the B-range , typically landing around 2.8-3.3 GPA territory. These aren’t suggestions—they’re institutional expectations designed to signal healthy academic rigor. When flagship CS courses start averaging 2.3 that’s a C+ for those keeping score , you’re looking at a system under stress. The kind of stress that happens when hundreds of students who’ve never struggled with math suddenly hit a wall that ChatGPT https://www.gadgetreview.com/chatgpt-becomes-fastest-app-to-hit-1-billion-users-as-claude-surges-in-ai-race can’t climb. The Perfect Storm of AI and Reality Heavy AI use, weak preparation, and strict enforcement created an academic collision course. Students arrived at Berkeley having used AI tools https://www.gadgetreview.com/ai-powered-websites-you-didnt-know-can-supercharge-your-productivity throughout high school for everything from calculus homework to coding projects. They looked great on paper—until professors started requiring in-person exams and implementing serious academic integrity enforcement. The gap between AI-assisted homework performance and actual mathematical reasoning became impossible to hide. Add understaffed courses and reduced office hours to the mix, and you get a generation of students who can orchestrate AI tools but struggle with linear algebra proofs when the laptop closes. When Grades Lose Their Meaning Researcher Igor Chirikov warns AI undermines the “informational value” of academic assessment. Igor Chirikov , whose research documented the national AI grade inflation trend, puts it bluntly: when students use AI on assignments, grades become inflated relative to actual knowledge. The result? Transcripts that don’t reflect real skills, creating a credibility crisis for academic assessment. Berkeley’s harsh correction https://www.dailycal.org/news/campus/academics/failing-grades-soar-as-professors-see-greater-ai-usage-dwindling-math-skills-in-uc-berkeley/article 16fad0bf-02cb-4b8c-8d88-888ffd9f8608.html reveals what happens when institutions try to restore meaning to grades in an AI-saturated world. Other universities inflating grades might avoid short-term student complaints, but they’re potentially graduating computer science majors who can’t debug code https://www.gadgetreview.com/the-20-most-frustrating-computer-problems-and-how-to-fix-them-fast without algorithmic assistance. Berkeley’s experiment offers a preview of higher education’s reckoning with artificial intelligence. The question isn’t whether AI will transform learning, but whether institutions can redesign assessment to capture genuine understanding. For now, Berkeley’s CS students are learning that in the real world of silicon and circuits, there’s no autocomplete for critical thinking.