UCLA’s Undergraduate Education office published a generative-AI citation rubric Friday that every syllabus template must link starting fall 2026, requiring students to tag paragraphs assisted by large language models and forbidding wholesale bibliography generation without human verification of each source—a policy Jason Kim covers as what actually changes inside lecture halls, not vendor keynote promises.
What the board document changes
The rubric does not ban AI; it creates four disclosure levels from “no AI use permitted” in closed-book labs to “AI brainstorming allowed, citations human-checked” in upper-division writing courses. Instructors pick a level on a standardized syllabus checkbox; registrars block grade submission until the box is set, mirroring how UCLA already enforces academic integrity statements.
Students must append an AI use log on essays over 1,500 words listing prompts and models. Paste-only Works Cited pages from chat tools count as fabrication if URLs fail resolver checks—a rule aimed at hallucinated journals TAs flagged last spring.
Which teachers and students, in which buildings
Life-science labs with prebuilt LMS quizzes see little change; humanities and social-science TAs in Powell Library writing sections bear the verification burden. International students on F-1 visas asked whether AI logs could trigger honor-code reports that affect status; the dean’s FAQ says process is unchanged but logs become discoverable evidence.
Implementation mess
United Auto Workers academic student employees said training modules dropped Friday without contracted prep time; UCLA Teaching and Learning promised asynchronous videos but no live workshops until October. Department chairs may override rubric levels for graduate seminars—a loophole that worries writing program directors who want consistent standards in entry-level courses.
Canvas plugins from three ed-tech vendors claim auto-detection; UCLA IT declined to certify them, pushing detection back onto human reading time TAs already lack.
Who is selling into schools
Turnitin and two startups emailed departments offering “rubric-aligned” scoring add-ons at per-seat pricing; the central office warned instructors those tools are optional purchases, not funded centrally. Library licensing for JSTOR and ProQuest unchanged—policy assumes students still touch paywalled PDFs.
Budget and time
No new line item appears in the published budget; workload shifts to existing TA hours. Kim’s question is whether add-drop week instructors will actually read logs or treat them as checkbox compliance—early pilot courses in political science reported 40 percent of logs copied boilerplate language without editing.
What happens after add-drop
Honor council cases this fall will test whether unlabeled AI paragraphs constitute plagiarism under the updated code; precedent from spring amnesty periods expires. Other UC campuses watch UCLA; Berkeley faculty senate scheduled a October compare-and-contrast hearing.
The building-level story is staffing: a rubric on a website does not grade papers—TAs with four sections and no detection software do, unless departments fund time the central policy assumes but does not pay for.
Transfer and AP students
Community college transfers arriving with different AI norms may pick courses before reading disclosure levels; UCLA advisers said they will add rubric icons to the online schedule by week two, too late for students who stacked writing-heavy loads in week one.
