AI Grading Tools Give Failing Scores to Correct Student Work, Discriminating Against Non-Native Speakers

In January 2025, multiple educational institutions across the United States faced scrutiny after AI-powered grading systems were documented systematically assigning failing or significantly reduced scores to academically sound student work. The incidents primarily affected students with non-standard writing styles, including non-native English speakers, students from diverse cultural backgrounds, and those with learning differences that influenced their writing patterns. Universities including Arizona State University, University of California system schools, and several large K-12 districts had implemented these AI grading tools to handle increased enrollment and reduce faculty workload. The discrimination became apparent when graduate student teaching assistants and faculty members began manually reviewing grades that seemed inconsistent with student performance in other assessments. In one documented case at a major state university, an ESL graduate student received a failing grade on a research paper that was later reviewed by three faculty members and deemed to be of B+ quality. The AI system had penalized the student for sentence structures and vocabulary choices that, while different from standard academic English, were grammatically correct and effectively communicated complex ideas. Similar patterns emerged across institutions, with AI systems showing particular bias against certain grammatical constructions common in Hispanic English, African American Vernacular English, and various international English dialects. The scope of the problem became clear when advocacy groups conducted systematic reviews of AI-graded assignments. Data analysis revealed that students with non-English surnames were 40% more likely to receive grades more than one letter grade lower than human reviewers assigned to identical work. International students, particularly those from Asia and Latin America, showed the most significant grade disparities. The AI systems appeared to prioritize narrow stylistic conventions over content accuracy, critical thinking, and substantive knowledge demonstration. Many affected students were unaware of the AI grading and had no mechanism to appeal algorithmic decisions. Educational institutions initially defended the AI systems, citing consistency and efficiency benefits. However, mounting evidence of systematic bias led to emergency reviews and temporary suspensions of automated grading at several major universities. Faculty senates at multiple institutions passed resolutions demanding transparency in AI grading algorithms and mandatory human oversight. The controversy intensified when affected students and advocacy groups filed class-action lawsuits alleging civil rights violations and educational discrimination. The incidents highlighted broader concerns about algorithmic bias in educational assessment and the lack of regulatory oversight for AI tools that directly impact student academic outcomes and future opportunities.