Stanford answered the AI question early. In February 2023, it adopted guidance that ties AI directly to the Honor Code. It's short, and it's the rule that applies whenever a syllabus is silent.
Here's what it says, what the campus AI Playground gives you, and which models fit Stanford coursework in CS, writing and design.
What Stanford actually says
We checked these pages in October 2026.
The Honor Code guidance. The Office of Community Standards publishes the generative AI policy guidance, adopted February 16, 2023. Absent a clear statement from the instructor, using generative AI is "treated analogously to assistance from another person." Using it to substantially complete an assignment or exam is not allowed. The example it gives of a prohibited use is entering exam or assignment questions into an AI tool. Instructors set their own course policies, and students should acknowledge any use beyond incidental use. When unsure, disclose and ask.
The AI Playground. University IT runs the Stanford AI Playground, open to students as well as faculty and staff. It offers models from several companies in one interface, and your data stays inside Stanford's environment. Since May 2026 it has been approved for high-risk data, including personal information, but personal health information is still not allowed in prompts or attachments. Some features, like one image agent, aren't yet available to students.
What that means in practice
- No AI statement means treat AI like a classmate. If you couldn't ask a friend to do it, don't ask a model.
- Pasting the question in is the clearest violation. The guidance names it directly.
- Disclose anything beyond incidental use. And when in doubt, disclose.
- The Honor Code still runs on trust. Stanford is known for unproctored exams. That trust is exactly why AI misuse on an exam is treated seriously.
Which models suit Stanford coursework
CS
Stanford's CS courses often set detailed rules on what help is allowed, and many treat AI like any other collaboration. Read them closely. Where AI is allowed for learning, ask for explanations, not code.
Try: "I don't understand why my linked list loses its head node after this insert. Ask me questions until I find the bug myself."
ChatGPT and Claude are both strong for code explanation. If you're comparing models for programming, see GPT-6 vs Claude Opus 5.5.
PWR and writing
The Program in Writing and Rhetoric teaches research-based argument. If your PWR section allows any AI, keep it to feedback on your own work.
Try: "Here's my research question and my three strongest sources. Which source is doing the least work for my argument, and what kind of source would fill the gap?"
Design and the d.school
Design thinking projects move fast through ideas. Where it's allowed and disclosed, image models help you sketch directions before you prototype. Nano Banana is good for editing an image by describing changes. FLUX is good for photorealistic concepts.
Research and the humanities
For literature searches, use a model that links real sources, then read them. Our Gemini vs ChatGPT for research students post compares two good options for working through long papers.
Try: "Summarize the main debate in this paper's literature review in five bullet points, and quote the sentence each point comes from."
Asking for quotes makes it easy to check the summary against the paper.
Gray areas worth thinking through
- Spelling and grammar tools. A spell checker is incidental use. An AI rewrite of a whole paragraph isn't, so it needs acknowledging, and only where the course allows it.
- Studying. Asking a model to explain a lecture concept is different from pasting in a problem set question. The guidance names the second one as prohibited.
- Group projects. Agree as a group on what AI use the course allows, and disclose it together. One person's undisclosed use becomes everyone's problem.
- Problem set hints. If you'd need to share the question to get the hint, treat it as entering assignment questions.
How to use AI at Stanford without breaking the rules
- Read each course's AI policy. If there isn't one, the Honor Code default applies.
- Never put exam or assignment questions into an AI tool unless the instructor has said you can.
- Disclose anything more than incidental use. A sentence at the end is usually enough.
- Use the AI Playground for Stanford material. It keeps data inside Stanford, and still no PHI.
- Verify every fact and source. UIT itself warns that models make errors.
- Ask before, not after. A quick question to the instructor or CA saves a conversation with Community Standards.
Where LazySusan fits
For coursework and anything involving Stanford data, the AI Playground is the right tool. It's provided by the university and keeps data inside Stanford.
LazySusan is for what's outside that: a startup idea, a personal project, a side gig or creative work that needs image, video and voice models next to chat. It brings ChatGPT, Claude, Gemini, Grok, Perplexity and DeepSeek into one workspace with one bill, with models like Veo, Runway and ElevenLabs alongside. Multi-Chat sends one prompt to several models and lines up the answers. Don't put Stanford data into it.
The Student plan is $99 a year with a .edu email, with 5,000 tokens a month that roll over. Prompt School covers prompting basics, and you can sign up here.
LazySusan isn't affiliated with or endorsed by Stanford University. The Honor Code and your syllabus come first.
The short version
At Stanford, unless your instructor says otherwise, AI counts as help from another person, and you can't use it to substantially complete an assignment or exam. Read each policy, disclose anything beyond incidental use, keep Stanford data in the AI Playground and verify everything.