← 政策图谱Atlas 申请环节Admissions 学生规范Guidelines
No.16 / 52 · 52 所大学 AI 使用政策比较研究No.16 of 52 · AI Use Policies at 52 Universities

UC Berkeley

University of California, Berkeley加州大学伯克利分校

来源层级 · 混合来源Source level · Mixed 9 / 12 项有明文规定9 of 12 provisions addressed 采集于 2026-08-21Accessed 2026-08-21 申请环节 · UC 系统共同规定Admissions · UC system (shared source)

ai.policy.nestudy.cn/Berkeley

原文摘录Excerpt

选取标准：对高中读者最具引用价值的一句原文。Selected as the sentence most quotable for a high-school reader.

GenAI should never be used in a way that would constitute plagiarism if the source were human.

↓ 在原文中的位置↓ Where it appears in the source

十二项条款The twelve provisions

9 / 12

每项附该校原文中的依据语句，以及图谱 52 校中持同一立场的校数。颜色越深，规定越明确、越具约束力（不代表优劣）；斜纹表示该来源未作表述。Each provision gives the supporting sentence and how many of the 52 universities in the atlas take the same position. Darker means more explicit and binding (not better); hatching means the source does not address it.

许可Permission使用许可由谁决定who decides whether AI may be used

默认规则Default rule

Default rule默认规则

作业未说明可否使用 AI 时，适用何种规则？When an assignment says nothing about AI, what applies?

默认禁止Prohibited by default主流口径MOST COMMON
If considering a new use of generative AI in studies or work, the user is responsible for considering the ethics and risks and must “obtain approval from your instructor/responsible unit head.”

编码说明CODING NOTEAcademic work not allowed by the instructor is prohibited use, and any new use of generative AI must have instructor or unit-head approval.

52 校中 19 所与此相同 · 该条主流口径：默认禁止（19/45）19 of 52 take this position · most common: Prohibited by default (19/45)

↓ 在原文中的位置↓ Where it appears in the source在图谱中与其他学校对照 ↗Compare with the others in the atlas ↗

课程规则优先Instructor override

Instructor override课程规则优先

任课教师的规定是否优先于校级规定？Does an instructor's rule take precedence over the institution's?

教师可定但有底线Instructor rules within limits主流口径MOST COMMON
Completion of academic work in a manner not allowed by the instructor.

编码说明CODING NOTEThe instructor defines what AI use is allowed in academic work, but the campus data and confidentiality prohibitions apply regardless of course rules.

52 校中 33 所与此相同 · 该条主流口径：教师可定但有底线（33/45）33 of 52 take this position · most common: Instructor rules within limits (33/45)

↓ 在原文中的位置↓ Where it appears in the source在图谱中与其他学校对照 ↗Compare with the others in the atlas ↗

分级许可Tiered scale

Tiered scale分级许可

是否设有分级的使用许可框架？Is there a graduated permission framework?

未作表述Not addressed

本项目采集的官方来源未就此条款作出表述；这不代表该校没有相关规定。The official source collected here does not address this provision; this does not mean the university has no rule on it.

52 校中 32 所同样未作表述32 of 52 are likewise silent

在图谱中与其他学校对照 ↗Compare with the others in the atlas ↗
透明Transparency使用后须说明的内容what must be declared after use

披露要求Disclosure

Disclosure披露要求

使用 AI 是否须予披露？为强制还是建议？Must AI use be disclosed, and is that required or recommended?

依课程而定Course dependent
When AI is permitted, instructors may require students to include an acknowledgment and share the prompts and outputs used at submission.

编码说明CODING NOTEAcknowledgment is something instructors may require, not a stated institution-wide duty.

52 校中 14 所与此相同 · 该条主流口径：强制披露（20/46）14 of 52 take this position · most common: Required (20/46)

↓ 在原文中的位置↓ Where it appears in the source在图谱中与其他学校对照 ↗Compare with the others in the atlas ↗

过程留证Process evidence

Process evidence过程留证

是否须保留草稿、提示词或编辑历史？Must drafts, prompts or edit history be kept?

建议留证Recommended主流口径MOST COMMON
When AI is permitted, instructors may require students to include an acknowledgment and share the prompts and outputs used at submission.

编码说明CODING NOTESharing prompts and outputs is presented as an option instructors may require, not a standing requirement.

52 校中 18 所与此相同 · 该条主流口径：建议留证（18/19）18 of 52 take this position · most common: Recommended (18/19)

↓ 在原文中的位置↓ Where it appears in the source在图谱中与其他学校对照 ↗Compare with the others in the atlas ↗
责任Accountability错误与违规的责任归属responsibility for errors and violations

核实责任Verification duty

Verification duty核实责任

AI 输出有误或虚构引注时，由谁负责？Who is responsible when AI output is wrong or a citation is invented?

未作表述Not addressed

本项目采集的官方来源未就此条款作出表述；这不代表该校没有相关规定。The official source collected here does not address this provision; this does not mean the university has no rule on it.

52 校中 14 所同样未作表述14 of 52 are likewise silent

在图谱中与其他学校对照 ↗Compare with the others in the atlas ↗

违规定性Integrity framing

Integrity framing违规定性

违规使用在纪律体系中如何定性？How is a violation classified in the disciplinary system?

明确等同抄袭Explicitly plagiarism
The guidance states that GenAI should never be used in a way that would constitute plagiarism if the source were human. Having another person or entity write a substantive part of an assignment can include using GenAI tools.

编码说明CODING NOTEFramed on the instructor-facing page; the compliance advisory instead treats academic work not allowed by the instructor as prohibited use.

52 校中 13 所与此相同 · 该条主流口径：学术不端（28/44）13 of 52 take this position · most common: Academic integrity violation (28/44)

↓ 在原文中的位置↓ Where it appears in the source在图谱中与其他学校对照 ↗Compare with the others in the atlas ↗

检测器证据Detector evidence

Detector evidence检测器证据

AI 检测工具的结果能否作为证据？Can AI-detection results be used as evidence?

不得作为唯一或主要证据Not sole or primary evidence主流口径MOST COMMON
The page also states that GenAI detection tools are not a viable or recommended enforcement option. Berkeley does not enable Turnitin’s AI detector because of privacy, security, accessibility, effectiveness, and actionability concerns.

编码说明CODING NOTEDetectors are called not viable or recommended for enforcement, and Berkeley does not enable Turnitin's AI detector.

52 校中 8 所与此相同 · 该条主流口径：不得作为唯一或主要证据（8/14）8 of 52 take this position · most common: Not sole or primary evidence (8/14)

↓ 在原文中的位置↓ Where it appears in the source在图谱中与其他学校对照 ↗Compare with the others in the atlas ↗
边界Boundaries数据、版权与工具的限制limits on data, copyright and tools

数据与隐私Data & privacy

Data & privacy数据与隐私

哪些内容不得输入 AI 工具？What must not be entered into an AI tool?

禁止上传受限数据Restricted data prohibited主流口径MOST COMMON
Unless specifically allowed, “no personal, confidential, proprietary, or otherwise sensitive information may be entered into or generated as output.”

编码说明CODING NOTENamed categories are barred as input and output unless specifically allowed; UC guidance warns uploaded content can be saved and reused by the vendor.

52 校中 20 所与此相同 · 该条主流口径：禁止上传受限数据（20/31）20 of 52 take this position · most common: Restricted data prohibited (20/31)

↓ 在原文中的位置↓ Where it appears in the source在图谱中与其他学校对照 ↗Compare with the others in the atlas ↗

机构工具Institutional tools

Institutional tools机构工具

学校是否指定或提供 AI 工具？Does the institution designate or provide AI tools?

未作表述Not addressed

本项目采集的官方来源未就此条款作出表述；这不代表该校没有相关规定。The official source collected here does not address this provision; this does not mean the university has no rule on it.

52 校中 28 所同样未作表述28 of 52 are likewise silent

在图谱中与其他学校对照 ↗Compare with the others in the atlas ↗
理由Rationale规则的依据the stated basis for the rules

学习理由Learning rationale

Learning rationale学习理由

规则是否以“不得绕过应习得的能力”为依据？Are the rules grounded in not bypassing the skills to be learned?

一般性表述General statement
Clear course policies should explain what is allowed and why, connect AI rules to learning goals, and support transparency, integrity, and shared responsibility.

编码说明CODING NOTEAdvises instructors to connect AI rules to learning goals; the source does not justify a specific rule by the skill being assessed.

52 校中 15 所与此相同 · 该条主流口径：明确以学习目标为据（22/37）15 of 52 take this position · most common: Explicit learning-objective rationale (22/37)

↓ 在原文中的位置↓ Where it appears in the source在图谱中与其他学校对照 ↗Compare with the others in the atlas ↗

补充信息Further detail

要求披露的内容What must be disclosed

提示词PromptsAI 输出内容AI output

申请环节的 AI 政策AI in the application

admissions/16-UC-Berkeley.md

该校招生办公室就申请人使用 AI 的公开表述，以本科新生申请为主。英文为逐字原文，中文为整理者的分析。证据状态描述找到了什么材料，不是宽严等级；未说明不代表允许。What this university's admissions office says about applicants using AI, chiefly for first-year undergraduate applications. English passages are verbatim; the Chinese is the collector's analysis. The evidence status describes what material was found, not how strict the rules are; silence does not mean permission.

证据状态Evidence status UC 系统本科 AI 规则/指导（共同来源）UC system undergraduate AI rules and guidance (a shared source, not a campus policy)

UC 系统申请诚信声明明确允许 AI 辅助可读性，要求本人独立写作及本人内容/最终文字；2026–27 指南另提早期构思/整理。非 Berkeley 独有。LL.M. 个人陈述有另行的生成式 AI 限制，仅限该项目。

52 所大学对照 →All 52 compared →

来源信息Source information

  • 访问日期Accessed 2026-09-24 (Asia/Shanghai)
  • 格式说明Format note Short English excerpts from live official webpages/PDFs; whitespace and layout normalized, wording retained. Chinese findings are editorial interpretation, not official quotations. Local text snapshots retain provenance; unavailable pages are explicitly marked.
  • 适用范围Scope note Primary: Berkeley undergraduate PIQs and campus application FAQ. S2 belongs to the UC SYSTEM (not a Berkeley-specific policy). S3 is graduate LL.M. only.
  • 证据状态Evidence status uc-system-ai-guidance — UC-SYSTEM application-integrity policy and first-year guidance verified; not a campus-specific rule. Separately scoped LL.M. policy also verified.
  • 相关校内／课程指导（范围不同）Related university/course guidance (different scope) Original file

中文结论（解读，非原文）

  • 本科结论必须拆开：Berkeley 自己的 PIQ 页面强调个人选择、收获和可辨识的个人信息，并未直接提到 AI；本科 FAQ 也未检出 AI 专条。[S1][S4]
  • UC 系统现行 2026–27 新生申请演示文稿第 67 页明确：AI 可以帮助早期构思或整理思路，但最终 PIQ 应是本人的作品和声音，且申请须同意完整性声明。此条作者是 UC 系统，不是 Berkeley。[S2-Q1][S2-Q2]
  • 研究生另记：Berkeley Law LL.M. 个人陈述必须是本人思想和文字，文中直接包含生成式 AI 限制；另有可选 Generative AI Questions，但不是个人陈述许可例外。[S3]

Official sources and exact excerpts

[S1] Personal Insight Questions - Office of Undergraduate Admissions

  • 发布单位Publishing unit University of California, Berkeley — Office of Undergraduate Admissions
  • 链接URL https://admissions.berkeley.edu/apply-to-berkeley/application-resources/personal-insight-questions/
  • 页面日期Page date Not stated on the retrieved page.
  • 适用对象Audience / scope undergraduate
  • 来源性质Authority official-advice
  • 采集时间Retrieved full-text; 2026-09-24T23:04:32+08:00
  • 范围与核验说明Scope / verification note Campus PIQ advice for first-year and transfer applicants; retrieved text does not mention AI.

[S1-Q1] Personal reflection expected in PIQs

Thoughtfully describe not only what you’ve done, but also the choices you have made and what you have gained as a result.

[S1-Q2] Grammar is not a separate evaluation criterion; not an AI permission

Your answers are not evaluated on correct grammar, spelling, or sentence structure, but these qualities will enhance overall presentation and readability.

[S1-Q3] Purpose of personal insight responses

If we have learned very little about you, your answers were not successful.

[S2] Presenting yourself on the first-year UC application

  • 发布单位Publishing unit University of California — UC Admissions (system-level counselor/applicant guide)
  • 链接URL https://admission.universityofcalifornia.edu/counselors/_files/documents/presenting-yourself-on-the-uc-application-first-year.pdf
  • 页面日期Page date Not stated; cover identifies the 2026–27 application cycle.
  • 适用对象Audience / scope undergraduate
  • 来源性质Authority official-advice
  • 采集时间Retrieved full-text; 2026-09-24T23:05:26+08:00
  • 范围与核验说明Scope / verification note SYSTEM authorship, not Berkeley- or UCLA-specific policy. First-year UC application presentation, live PDF page 67. This quotes the current PDF, not longer/older search-index snippets. The full UC Statement of Application Integrity is researched separately by the parent project.

[S2-Q1] UC-system first-year PIQ advice, PDF page 67

Can be a useful tool for brainstorming or organizing your thoughts in the early stages, but your final PIQ responses should be your own work and voice

[S2-Q2] System guide references the required application-integrity statement

All applicants must agree to UC’s Statement of Application Integrity when submitting their UC application!

[S3] Ready to Apply - UC Berkeley Law

  • 发布单位Publishing unit UC Berkeley School of Law — LL.M. Admissions
  • 链接URL https://www.law.berkeley.edu/admissions/llm-admissions/application-admission/application-checklist/
  • 页面日期Page date Not stated on the retrieved page.
  • 适用对象Audience / scope graduate
  • 来源性质Authority policy
  • 采集时间Retrieved partial; 2026-09-24T23:06:20+08:00
  • 范围与核验说明Scope / verification note Graduate LL.M. program only. Standard web_fetch returned readable HTTP 200; local snapshot retains the relevant sections, not the entire page. The separate optional Generative AI Questions are not evidence that AI may write the personal statement.

[S3-Q1] LL.M.-only personal-statement authorship and AI restriction

The thoughts and words contained in your personal statement must be your own and no one else should assist in its creation beyond basic proofreading or critiquing. This must be your original work and should be written without assistance from others, including generative AI (e.g., ChatGPT, Claude, etc.).

[S3-Q2] Separate optional components; not a personal-statement exception

There are two optional opportunities to provide additional information as part of the application: the Video & Written Assessment and the Generative AI Questions. Applicants may choose to complete either or both components. Neither is required to submit a complete application.

[S4] Application FAQs - Office of Undergraduate Admissions

  • 发布单位Publishing unit University of California, Berkeley — Office of Undergraduate Admissions
  • 链接URL https://admissions.berkeley.edu/application-faqs/
  • 页面日期Page date Updated February 2026
  • 适用对象Audience / scope undergraduate
  • 来源性质Authority official-advice
  • 采集时间Retrieved full-text; 2026-09-24T23:07:15+08:00
  • 范围与核验说明Scope / verification note Application FAQ checked as a campus-policy fallback; no AI-specific applicant instruction found in the retrieved FAQ text.

No policy quotation: this source is retained only for search-scope or retrieval documentation.

UC 共享系统规则补证（主审整合）

[S5] Statement of Application Integrity

  • Publishing unit: University of California — undergraduate application system
  • Official URL: https://apply.universityofcalifornia.edu/docs/StatementOfIntegrity.pdf
  • Page date: Last updated August 2023（PDF 页末明示；不是 2026 更新）
  • Retrieved: 2026-09-24T15:10:13.275383+00:00; full-text; public PDF actually downloaded
  • Scope / authority: undergraduate-system-wide; system-application-policy
  • 归属说明：这份文件适用于 UC 共享本科申请，不是本校单独制定的 AI 政策。并列保留 2026–27 申请指南，正式诚信声明与指导材料不得混为同一版本。

[S5-Q1]

The personal insight responses should reflect the student’s own ideas and experiences, be independently written by the student and reflect the student’s personal writing style. Students may receive advice on content and editing, including the use of generative artificial intelligence software to assist with readability, but content and final written text must be their own.

[S5-Q2]

UC conducts regular screenings to verify the integrity of the responses, may request authentication of the content or writing as the student’s, and will take action when it is determined that the integrity of the response is compromised.

[S5-Q3]

Students found to have falsified application information will be subject to penalties up to and including denial of admission, withdrawal of an admission offer, registration cancellation, expulsion or revocation of a University of California degree.

中文解读：正式声明允许内容与编辑建议，包括生成式 AI 的可读性辅助；但 PIQ 须体现本人思想、经历和文风，由本人独立写作，内容和最终文字须为本人所有。UC 可要求真实性核验；引文中的处分针对虚假申请信息，不能说成“任何 AI 使用必然受同样处分”。

使用边界（中文解读，不是学校原话）

  • 本科—系统指南提到的辅助：早期构思、整理自己的思路；最终 PIQ 保持本人作品和声音。[S2-Q1]
  • 本科—边界：不能把系统的辅助建议写成允许 AI 代写最终 PIQ；Berkeley 对拼写/语法不单独评分也不等于授权 AI 改写。[S1-Q2][S2-Q1]
  • 本科—未单列：AI 翻译、研究附件、结构或措辞重写阈值、披露要求和处分细节。系统声明的可读性辅助、核验与后果见上方补证；未说明的翻译、具体重写阈值仍不推断。
  • 仅 LL.M.：个人陈述原文限制超出基础校对/评论的他人协助，并明示应不借助生成式 AI 写作；不要据校对措辞自行扩张 AI 许可。[S3-Q1]

限制与未说明事项

  • 本科 AI 证据来自 UC 系统作者，不是 Berkeley 校园独有声明。学校 status 必须结合此归属说明理解。
  • 校园 PIQ 页与 Application FAQs 未检出 AI 专条，只是本次公开检索范围内的结果，不能证明不存在其他规则。
  • 系统 PDF 为 first-year 2026–27 版本；不自行把该页扩大为所有转学/专项申请规则。
  • 主审已补入公开 UC 申请诚信声明的原文摘录，见“UC 共享系统规则补证”；未登录当季申请界面核对最终认证流程。
  • AI 翻译、研究补充材料、披露和具体编辑阈值未在所引本科条目逐项说明；一般虚假申请后果见补入的系统声明，不能外推为任何 AI 使用的自动处分。
  • Berkeley Law LL.M. 限制及其可选 AI 题目只适用于该研究生项目；不得外推本科或所有 Berkeley 研究生。
检索记录（2026-09-24）
  • site:berkeley.edu admissions "AI" "personal insight" — 检索本科 PIQ 与校级 AI 线索；校园教学内容不作申请证据。
  • site:admissions.berkeley.edu "artificial intelligence" OR "ChatGPT" application — 未得到经核验的 Berkeley 独有本科 AI 条款；搜索结果中项目申请与其他大学材料不能替代。
  • site:admissions.berkeley.edu "application" "authentic" "AI" — 结果相关性差，未据此认定无政策；继续直接检查官方本科材料。
  • Berkeley undergraduate admissions FAQ artificial intelligence essays authentic own work — 第三方汇总指向 UC 系统规则，仅作为线索；直接取得 Berkeley PIQ 页与系统指南。
  • site:admissions.berkeley.edu "FAQs" application — 定位并完整读取 Application FAQs；未检出 AI、ChatGPT 或 artificial intelligence。
  • site:admission.universityofcalifornia.edu "AI" "personal insight" — 官方系统 PDF 搜索摘要与现场版本不同；直接下载的 2026–27 PDF 第 67 页才是本页引文依据。
  • Direct graduate cross-check: Berkeley Law LL.M. Ready to Apply — 普通 requests 为 403；标准 web_fetch 成功 HTTP 200，核验个人陈述及独立可选组件；仅作研究生补充。

本页仅报告上述范围内已核验的材料；未找到、未明确说明或抓取失败均不等于允许使用 AI。课程/在校生政策不自动适用于申请者。

官方原文The official text

univ/16-UC-Berkeley.md

本项目采集的该校官方文本。图谱引用的语句已高亮，句末标签注明对应条款，点击可返回该条款。The official text collected for this university. Sentences the atlas cites are highlighted; the tag after each names the provision it supports and links back to it.

来源信息Source information

  • 访问日期Accessed 2026-08-21
  • 格式说明Format note Direct excerpts and concise, source-grounded summaries from official UC Berkeley webpages; navigation removed. Directly quoted wording is retained.
  • 适用范围Scope note The OERCS page is a campus risk/compliance advisory. The Center for Teaching and Learning page is instructor-facing guidance; together they do not constitute one universal course policy.
  • 用途说明Use note Source material for comparison and later high-school adaptation.

Appropriate Use of Generative AI Tools

The advisory supports innovation while protecting institutional, personal, and proprietary information.

Prohibited use

  • “Completion of academic work in a manner not allowed by the instructor.课程规则优先Instructor override”
  • Unless specifically allowed, “no personal, confidential, proprietary, or otherwise sensitive information may be entered into or generated as output.”数据与隐私Data & privacy
  • Protected examples include FERPA student records, non-public instructional materials, proprietary or unpublished research, and information classified at higher protection levels.知识产权IP & copyright

If considering a new use of generative AI in studies or work, the user is responsible for considering the ethics and risks and must “obtain approval from your instructor/responsible unit head.”默认规则Default rule

UC legal guidance quoted by Berkeley warns that, unless specific precautions are taken, “any content that is provided to generative AI tools can be saved and reused by the company offering the tool.”

Course policy and acknowledgment

Clear course policies should explain what is allowed and why, connect AI rules to learning goals, and support transparency, integrity, and shared responsibility.学习理由Learning rationale When AI is permitted, instructors may require students to include an acknowledgment and share the prompts and outputs used at submission.披露要求Disclosure过程留证Process evidence

The guidance states that GenAI should never be used in a way that would constitute plagiarism if the source were human.摘录Excerpt Having another person or entity write a substantive part of an assignment can include using GenAI tools.违规定性Integrity framing

The page also states that GenAI detection tools are not a viable or recommended enforcement option. Berkeley does not enable Turnitin’s AI detector because of privacy, security, accessibility, effectiveness, and actionability concerns.检测器证据Detector evidence

以下为整理者所写的高中适用要点，非该校原文，不作为任何条款的依据。Below: the collector’s notes on application to high school — not the university’s text, and not used as evidence for any provision.

High-school adaptation takeaways

  • Teacher permission overrides general availability of a tool.
  • Never enter student records, unpublished work, private teaching materials, or other sensitive data into unapproved AI systems.
  • Require a short acknowledgment and, where useful, the prompts and relevant outputs.
  • Do not use AI detectors as automatic proof of misconduct.