Public systems and ten institutional cases
California Institution Findings
California’s public systems and selected institutions differ materially in formation purpose, faculty or parent authority, artificial-intelligence governance, privacy, assessment and protected independent work. The record supports preliminary findings—not a final combined ranking.
Project, Research, Analysis, Editorial by Patrick Russell; assisted by ChatGPT (OpenAI).
State guidance community colleges state universities and the University of California
Finding
California’s four principal public education systems have moved beyond ignoring artificial intelligence. Each now acknowledges risks to privacy, academic integrity, equity, authorship, and human judgment. The California Department of Education has produced the most explicit school-level model policy; the California Community Colleges use the clearest systemwide language that technology should amplify rather than replace human capacity; the California State University has pursued the broadest immediate access strategy; and the University of California has produced the strongest faculty-led analysis of agency, intellectual work, trustworthiness, and the right to refuse AI.
Yet none currently demonstrates a complete public model for developing a well-formed mind. Their strongest documents are frameworks, recommendations, pilots, access programs, or governance principles. The systems generally do not require a common definition of the human capacities education must protect, a verified baseline of unaided student performance, or comparable evidence that AI improves knowledge and judgment rather than convenience and output.
The structural problem
Public systems divide authority among state officials, governing boards, system offices, campuses, academic senates, faculty, districts, schools, vendors, families, and students. This distribution can protect local judgment and academic freedom. It can also disperse responsibility so thoroughly that no actor is accountable for the full formative outcome.
Evidence tableView 4 recordsHide 4 records
System | Primary role | AI posture | Central limitation |
|---|---|---|---|
California Department of Education | State guidance for transitional kindergarten through grade twelve; standards, resources, and model policy. | Human-centered, age-aware, privacy and integrity focused; local adoption remains voluntary. | Cannot show implementation merely by publishing guidance. |
California Community Colleges | System support across 116 locally governed colleges. | Human-centered AI, equity, pilots, faculty leadership, workforce preparation. | System principles may vary widely in course-level practice. |
California State University | Large teaching-oriented public university system with campus and faculty governance. | Broad tool access, training, partnerships, and rapid integration. | Access and adoption have moved faster than public evidence of learning gains. |
University of California | Research university system with strong faculty authority over curriculum. | Responsible AI principles, risk assessment, shared governance, agency, adaptability, trustworthiness. | Campus and disciplinary autonomy can produce inconsistency and fragmented student expectations. |
California Department of Education
What the new guidance gets right
The state’s updated Guidance for the Safe and Effective Use of Artificial Intelligence in California Public Schools explicitly says AI should strengthen human learning, creativity, and purpose rather than replace them. It addresses human-centered AI, literacy, access, academic integrity, and privacy. The companion model policy requires disclosure, original student work, limits on AI detection, authorized tools, consent when AI is used for grading or feedback, and the principle that AI should not substitute for critical thinking or required academic work.
- It distinguishes learning about AI from learning with AI.
- It preserves educators as essential to creativity, authentic voice, judgment, and healthy human connection.
- It recognizes bias, hallucination, privacy, academic integrity, well-being, and unequal access.
- It allows assignment-level discretion but expects the purpose and rules to be communicated to students and families.
- It rejects AI-detection output as the sole basis for discipline and calls for alternative evidence of authorship.
Why it remains incomplete
The guidance and model policy are expressly exemplary and nonmandatory. Local educational agencies may adapt them or decline to follow them except where separate law applies. This respects local control, but it also means California has not established a consistent minimum protection for unaided reading, writing, memory, calculation, reasoning, oral defense, or moral judgment.
Evidence tableView 5 recordsHide 5 records
Well formed mind test | Evidence in state materials | Gap |
|---|---|---|
Truth and verification | Critical evaluation, misinformation, bias, hallucinations, citation, and disclosure are addressed. | No statewide measure shows students become better at distinguishing truth from plausible output. |
Independent work | The model policy says AI must not replace critical thinking or required academic work. | It does not define the minimum work every student must perform unaided. |
Parent authority | The model includes communication, parent access to criteria, and consent for AI grading or feedback. | Local discretion can produce uneven notice, inspection, consent, and challenge procedures. |
Attention and relationships | Human connection, authentic voice, well-being, and educator judgment are recognized. | No clear protection against dependency, distraction, reduced stamina, or substitution of chatbot relationships. |
Accountability | Working groups, public meetings, guidance, and model policy are visible. | Adoption, classroom use, vendor compliance, and learning effects are not yet reported comparably. |
California Community Colleges
Strongest public system language on human capacity
The Chancellor’s Office states that technology should amplify human capacity rather than replace it. Its systemwide approach connects AI with equity, accountability, student success, faculty leadership, internal expertise, ethical use, governance, and evaluation. Current initiatives include faculty-led tool creation, AI course assistants, common data infrastructure, AI fellows, ambassadors, and professional development.
This is a more mature institutional stance than treating AI merely as software procurement. It recognizes that teaching, student support, governance, professional judgment, and system capacity must develop together.
The unresolved formation question
The system’s stated goals remain dominated by access, student success, scalability, and workforce preparation. Those goals are legitimate, especially for working adults and students historically excluded from higher education. But they do not themselves define the quality of mind a graduate should possess.
Evidence tableView 5 recordsHide 5 records
Strength | Risk | Evidence required |
|---|---|---|
Human-centered HUMANS principles and explicit nonreplacement language | The phrase can become aspirational if assignments do not protect independent thought. | Course rules, unaided assessments, oral defenses, student work, and learning comparisons |
Faculty fellows, ambassadors, and locally generated tools | Innovation may remain concentrated among volunteers and early adopters. | Participation, course reach, discipline coverage, and skeptical faculty perspectives |
AI course-assistant pilots | Engagement and response speed may be mistaken for understanding. | Pre and post learning, transfer, dependency, accuracy, persistence, and subgroup outcomes |
Industry partnerships and workforce alignment | Vendor interests may shape curriculum, tools, credentials, and definitions of readiness. | Contracts, governance, exit rights, data terms, faculty control, and employer outcome evidence |
Equity and access | Universal availability can broaden opportunity but also universalize weak pedagogy or surveillance. | Accessibility testing, privacy audit, cost, alternatives, and opt-out experience |
California State University
The broad access strategy
The California State University has pursued large-scale access to generative AI tools, resources, and training through a public-private initiative. The system describes the effort as support for teaching, learning, research, operations, student success, and California’s workforce. Academic Senate materials also show faculty concern about systemwide tool selection, faculty participation, academic freedom, integrity, privacy, intellectual property, workload, and governance.
The institutional advantage is reach: a systemwide program can reduce unequal access and provide common training and security arrangements. The danger is that procurement creates momentum before the educational purpose, course-level boundaries, and outcome measures are settled.
Assessment under the formation framework
This table applies the formation criteria to the California public record while keeping preliminary judgments separate from their evidence status.
Evidence tableView 8 recordsHide 8 records
Criterion | Preliminary judgment | Evidence status |
|---|---|---|
Access and affordability | Systemwide availability may reduce tool-access inequality. | Institutional claim with structural support |
Teacher and faculty preparation | Training and resources are central to the initiative. | Participation and practice not yet verified |
Independent work | Campus and course rules may preserve unaided work, but no common public minimum was established in this review. | Missing systemwide evidence |
Truth and verification | AI literacy and responsible use are discussed; actual verification requirements vary. | Fragmented evidence |
Faculty authority | Senate resolutions indicate active concern about consultation and selection. | Documented governance dispute |
Vendor governance | Partnership gives scale but raises dependence, data, interoperability, cost, and exit questions. | Contract evidence required |
Learning outcomes | The initiative promises educational and workforce benefits. | Comparable gains not yet demonstrated |
Formation purpose | Student success and workforce readiness are clear; a comprehensive account of judgment, character, and human flourishing is not. | Partial institutional position |
University of California
The strongest internal critique
The University of California Academic Senate’s December 2025 report is the most intellectually candid of the four system documents reviewed. It acknowledges that administrative technology units often moved before substantial faculty governance; that little policy work had been completed in academic integrity, personnel review, and data ownership; and that enthusiasm for research uses can coexist with concern about effects on reading, writing, creativity, and pedagogy.
The report organizes its recommendations around agency, adaptability, and trustworthiness. It defends faculty authority, discipline-specific rules, opt-in choices, the right to refuse some AI tools, transparency when instructors use AI to teach or evaluate, protection of student data, and consultation before procurement.
Human intellectual work
Most important for this investigation, the report states that learning remains active and demanding when AI is present and that generative tools cannot replace the intellectual work learning requires. It recommends assignments built around iteration, reflection, comparison, critique, revision, research, and argumentation. It also treats AI as an opportunity to ask what counts as knowledge, evidence, argument, authorship, and judgment.
Evidence tableView 5 recordsHide 5 records
Strength | Why it matters | Remaining gap |
|---|---|---|
Faculty agency and shared governance | Places pedagogical judgment with those responsible for disciplines and assessment. | Students may face inconsistent rules across courses and campuses. |
Right of refusal and opt-in tools | Recognizes that adoption is not automatically educational progress. | Practical alternatives and protections require implementation evidence. |
Trustworthiness and disclosure | Extends transparency obligations to institutional and faculty AI use, not only students. | Enforcement, vendor auditing, and student remedies remain to be tested. |
Active demanding learning | Directly rejects intellectual outsourcing and recommends process-rich assignments. | No minimum systemwide unaided performance standard. |
AI literacy in early coursework | Could establish a common vocabulary of limits, evidence, and ethical use. | Curriculum, reach, assessment, and learning outcomes remain future work. |
Comparison against the formation benchmarks
This table compares California’s principal public systems against the same formation benchmarks so differences in authority, policy and evidence remain visible.
Evidence tableView 7 recordsHide 7 records
Test | CDE | Community Colleges | CSU | UC |
|---|---|---|---|---|
Clear human end | Human learning, creativity, purpose, connection | Human capacity, equity, success, work | Student success and workforce readiness | Research, instruction, agency, trust |
Protected unaided work | Principle stated; minimum undefined | Not established systemwide | Not established systemwide | Faculty discretion; minimum undefined |
Parent or student authority | Parent notice and some consent in model policy | Student participation and access emphasized | Student access emphasized | Consent, transparency, and student data concerns |
Truth and verification | Strong guidance language | Ethical AI literacy | Responsible use and training | Knowledge, evidence, hallucination, disclosure |
Moral judgment | Ethical decision-making noted | Public values and ethics noted | Not a central systemwide formative end | Ethical, societal, legal, and disciplinary judgment |
Implementation evidence | Not comparable statewide | Pilots and programs underway | Large initiative underway | Decentralized and evolving |
Outcome evidence | Not sufficient for final score | Not sufficient for final score | Not sufficient for final score | Not sufficient for final score |
Relation to Mother of Divine Grace and Thomas Aquinas College
The contrast is not simply public versus private. It is coherence versus distribution. Mother of Divine Grace and Thomas Aquinas College begin with a defined formative purpose and design a curriculum around it. California’s public systems begin with multiple legal missions, constituencies, disciplines, access obligations, and governance layers. Their pluralism and scale are real strengths, but they make it harder to identify a single educational end and easier to treat policy compliance or technology access as evidence of formation.
The public systems should not imitate a religious curriculum. They can, however, adopt the structural lesson: identify the human capacities that must never be delegated, require repeated practices that cultivate them, preserve visible unaided performance, and measure whether graduates actually possess them.
COVID-era disruption remains important context for interpreting student readiness, recovery and the transition to artificial intelligence. Read the full mental-fitness, wellness and COVID analysis.
Required public system standard
A California standard compatible with the well-formed mind should be technologically open but educationally firm. It should permit institutions and disciplines to choose different tools while requiring each to protect and measure the same core human capacities.
Evidence tableView 9 recordsHide 9 records
Required element | Operational requirement |
|---|---|
Human purpose | State what education is forming beyond completion, employability, access, or satisfaction. |
Unaided baseline | Require recurring demonstrations of reading, memory, calculation, writing, oral explanation, creation, and judgment without AI. |
Purpose-bound AI | Permit a tool only when its educational purpose and the protected human work are explicit. |
Truth protocol | Require source verification, uncertainty, error analysis, disclosure, and correction. |
Human relationships | Protect teacher judgment, peer discussion, mentorship, family transparency, and access to a responsible person. |
Privacy and governance | Authorize tools, minimize data, disclose vendors, audit contracts, preserve alternatives, and define exit rights. |
Mental fitness | Monitor attention, perseverance, agency, belonging, and dependency without converting every difficulty into diagnosis. |
Evidence of benefit | Measure durable knowledge, transfer, independent judgment, equity, and long-term outcomes against a credible comparison. |
Public accountability | Publish adoption, use, cost, incidents, results, disagreements, and corrections in comparable form. |
Next evidence requests
- For CDE: track which local agencies adopt the model policy, what they change, and whether parent notice, consent, unaided work, and outcome reporting follow.
- For Community Colleges: obtain HUMANS criteria, pilot designs, vendor agreements, participating courses, learning measures, subgroup outcomes, and independent evaluations.
- For CSU: obtain system and campus contracts, privacy reviews, faculty consultation records, course-level rules, training reach, costs, opt-out alternatives, and learning results.
- For UC: track implementation of the Academic Senate recommendations, campus procurement, AI-literacy curricula, assignment repositories, disclosure rules, and student remedies.
- For all systems: compare pre-COVID, closure-period, recovery-period, and current evidence on learning, attention, attendance, persistence, relationships, wellness, and independent work.
Sources
California Department of Education AI guidance
California Department of Education model AI policy
California Department of Education AI working group
California Community Colleges AI system overview
California Community Colleges guidance and policy
California State University AI Commons
University of California AI Council
University of California Academic Senate AI report
Evidence note: official system materials establish policies, principles, programs, recommendations, governance positions, and reported activities. They do not by themselves verify classroom implementation or causal improvement in learning and judgment.
California Institution Findings
Educational formation artificial intelligence governance authorship and the evidence still needed
Main finding
The ten institutions reveal two different strengths. Several universities publish substantial artificial intelligence infrastructure, tool guidance, privacy rules, and course-policy resources. The formation-centered and project-centered institutions state more clearly what students should become and protect more visible forms of direct intellectual work. No institution yet supplies a complete public chain from educational objective to actual tool use, protected unaided performance, delayed learning, and student wellness.
The findings therefore do not produce a single ranking. They identify where the published record is strong, where a practice is partly visible, and where evidence remains unresolved. A missing public document does not prove that a policy or practice does not exist. A published policy does not prove implementation, and a completed assignment does not prove durable learning.
Cross institution determinations
This table consolidates the questions that cut across the institutional cases and records the determinations the shared evidence supports.
Evidence tableView 5 recordsHide 5 records
Question | Determination |
|---|---|
Strongest formation architecture | Thomas Aquinas College and Mother of Divine Grace describe the clearest integrated educational purposes and sequences |
Strongest public AI governance examples | Berkeley Law and Caltech publish restrictive rules directly tied to learning and authorship; Stanford UCLA USC and Loyola Marymount publish broader guidance and support |
Strongest authentic performance model | Thomas Aquinas College protects oral written mathematical and laboratory work; High Tech High emphasizes projects critique and public exhibition |
Strongest family role | Mother of Divine Grace places continuing instructional responsibility in the family; the other K through 12 public record is less complete on AI-specific parent authority |
Weakest common evidence area | Comparable delayed unaided learning and longitudinal wellness evidence |
How the findings should be read
The institutions differ in age range, mission, governance, selectivity, curriculum, residential life, family role, and legal setting. The comparison uses common human-formation questions while preserving separate standards for K through 12 and higher education. It does not treat a college seminar, a home lesson, and a charter-school project as interchangeable.
Evidence classifications
This table defines the strength and limits of each evidence classification so published rules, actual practice and verified outcomes are not confused.
Evidence tableView 5 recordsHide 5 records
Classification | Meaning | What it does not establish |
|---|---|---|
Published | An official current source states the objective rule program or tool arrangement | Consistent implementation or benefit |
Partly established | The source describes a practice but key details artifacts scope or authority remain incomplete | The complete operating condition |
Unresolved | The needed evidence was not located in the public record reviewed | That the policy practice or result does not exist |
Reported result | An institution or participant reports an experience or outcome | Causation prevalence or transfer |
Measured result | A defined measure and comparison are available | Durability unless the interval and removal condition support it |
Common questions
- What human capacity does the institution say education should form?
- Which artificial intelligence uses are allowed restricted or prohibited and by whose authority?
- Which accounts are approved and what happens to prompts student records and derived data?
- What work must students perform without generation and how is authorship verified?
- What can students do after assistance is removed and after time has passed?
- How are age consent disability family capacity relationships and wellness addressed?
Review date
The official-source record was reviewed through September 13 2026. Policies can change by institution, division, school, department, course, and term. The continuing monitoring process records meaningful published changes but does not infer classroom practice from webpage revision alone.
Higher education findings
The higher-education track includes seven institutions. Faculty and departmental authority matter because the mental act being taught differs across law, writing, computation, laboratory science, seminar discussion, and professional practice. Campus-wide tool access or general guidance should not erase those differences.
Evidence tableView 7 recordsHide 7 records
Institution | Most established strength | Principal unresolved question |
|---|---|---|
UC Berkeley | Central AI resources plus a restrictive Berkeley Law policy effective summer 2026 | Whether other units protect and measure independent reasoning in comparable ways |
UCLA | Six teaching principles literacy module faculty fellows and approved-tool guidance | Whether literacy and adoption change unaided student performance |
Stanford University | Student-facing learning guide course-specific authority and student policy deliberation | Whether proposed understanding checks became recurring assessed practice |
University of Southern California | Managed ChatGPT Edu access general policy and faculty control | Whether broad access improves learning beyond assisted production |
California Institute of Technology | Clear instructor authority disclosure privacy and skepticism about detector use | Whether protected transfer is measured across scientific and humanistic work |
Loyola Marymount University | AI-free and integrated options protected tools and attention to alternatives | How consistently choice and independent performance operate across courses |
Thomas Aquinas College | Unified curriculum direct discussion and a restrictive student rule | How the rule handles accessibility exceptions and later responsible AI competence |
Higher education judgment
The public universities and large private universities are building governance and access capacity faster than they are demonstrating learning. Thomas Aquinas College protects more human-only work but publishes less evidence about accessibility, exceptions, implementation reliability, and AI competence after graduation. The strongest future model would join explicit formation, faculty authority, managed tools, protected assessment, and delayed unaided outcomes.
UC Berkeley
Berkeley combines a campus AI hub and teaching resources with decentralized instructional authority. Berkeley Law now provides the clearest unit-level restriction in the Berkeley record reviewed.
Evidence tableView 5 recordsHide 5 records
Evidence area | Published record | Current judgment |
|---|---|---|
Educational purpose | Campus resources emphasize responsible use. Berkeley Law links its restriction to independent legal thought. | Published at the law school level; a comparable campus-wide formation standard is unresolved. |
AI rule | Berkeley Law policy effective summer 2026 restricts using AI to conceptualize draft revise translate or edit credited work unless an exception applies. | A strong protected-work rule for one professional school. Course exceptions and enforcement evidence remain incomplete. |
Tools and privacy | The AI Hub directs users to licensed tools training policies and risk guidance. | Approved account settings retention alternatives and unit-level practice require verification. |
Independent performance | The law policy protects student reasoning and authorship by limiting generative assistance. | No common public oral defense delayed transfer or cross-campus removal measure was located. |
Outcomes and wellness | Research and pilot activity are visible across Berkeley. | No public causal evidence links campus AI policy to durable learning trust workload or wellness. |
Preliminary finding
Berkeley demonstrates that a large university can combine central infrastructure with stricter unit rules. The evidence supports credit for governance and protected legal work, not a campus-wide claim of improved learning.
Recommended investigation
Preserve one complete assignment chain from a restrictive and a bounded-use course. Compare drafts, source verification, oral explanation, delayed transfer, workload, accessibility, and student understanding of the rule.
Principal official sources [1] [2] [3]
UCLA
UCLA has moved from general discussion toward an institutional teaching framework with six principles, a student literacy module, faculty fellows, workshops, an advisory committee, and an approved-tool guide.
Evidence tableView 5 recordsHide 5 records
Evidence area | Published record | Current judgment |
|---|---|---|
Educational purpose | The teaching framework names equitable access student literacy faculty support assessment integrity innovation and research-led adoption. | The principles guide direction but do not define a complete account of formation or a common unaided outcome. |
AI rule | UCLA supports course-specific policies that can allow limit or prohibit use. | The operating rule must be verified at course and assignment level. |
Tools and privacy | The official guide lists approved enterprise tools including Copilot Gemini ChatGPT Enterprise NotebookLM and Bedrock models. | Approval does not establish developmental fit actual configuration student choice or learning benefit. |
Independent performance | Assessment integrity is an explicit institutional principle. | No recurring university-wide post-assistance writing oral explanation or transfer check was located. |
Outcomes and wellness | The center states that adoption decisions should be driven by research. | Public evidence of delayed learning dependency belonging or wellness effects remains unresolved. |
Preliminary finding
UCLA has one of the most coherent campus frameworks in the group. Its next evidentiary step is to show whether literacy training and approved access change how students think and what they can do independently.
Recommended investigation
Evaluate the literacy module and selected Faculty Fellows courses using baseline work, actual-use records, disclosed assistance, protected assessment, delayed transfer, subgroup analysis, and student experience.
Principal official sources [4] [5] [6]
Stanford University
Stanford publishes unusually direct student guidance. It asks whether AI deepens engagement or delegates the learning process, treats substantial completion as prohibited, and preserves instructor-specific course rules.
Evidence tableView 5 recordsHide 5 records
Evidence area | Published record | Current judgment |
|---|---|---|
Educational purpose | The student guide centers informed choice critical evaluation accurate sourcing and human support. | Strong guidance does not establish adoption or assessed student competence. |
AI rule | Students are told to assume use is not allowed unless a course says otherwise and to disclose permitted assistance. | Actual permissions vary by course and must be preserved with each assignment. |
Student participation | More than 1200 students in over 70 COLLEGE 102 sections developed AI-era policy proposals in 2026. | The exercise establishes civic inquiry and participation; the winning proposal was not itself university policy. |
Independent performance | Students proposed individual understanding checks including oral and supervised verification. | Public evidence of recurring implementation and measured effect remains unresolved. |
Relationships and wellness | The guide warns that lifelike interaction cannot provide holistic support and directs students to people. | Longitudinal evidence on help seeking dependence belonging and attention is not public. |
Preliminary finding
Stanford offers the strongest public student-facing explanation of the difference between using a tool and delegating learning. Its most important assessment protection remains uneven because course policy and implementation are decentralized.
Recommended investigation
Follow the COLLEGE 102 proposals into actual courses. Document which understanding checks were adopted, how they affected behavior and workload, and whether students retained knowledge after AI-supported work.
Principal official sources [7] [8] [9]
University of Southern California
The University of Southern California created a university-scale managed-tool case by rolling out ChatGPT Edu in spring 2026 while retaining faculty control over classroom policies.
Evidence tableView 5 recordsHide 5 records
Evidence area | Published record | Current judgment |
|---|---|---|
Educational purpose | Provost guidance presents managed AI as a resource for teaching research and work under university requirements. | Account availability is not a theory or measure of learning. |
AI rule | Faculty remain responsible for course permissions and students remain responsible under academic-integrity requirements. | Consistency across schools courses and assignments requires direct evidence. |
Tools and privacy | ChatGPT Edu provides a USC-managed environment and the general AI policy requires accuracy review and responsible use. | Configuration retention integration permissions and meaningful alternatives require verification. |
Independent performance | Faculty guidance encourages policy clarity and assessment planning. | No public university-wide removal test or supervised post-assistance standard was located. |
Programs and outcomes | USC has expanded AI research education and professional programs. | Program growth and research output do not establish improved student learning or preserved voice campus-wide. |
Preliminary finding
USC is the clearest scale-and-governance case. The managed environment improves institutional control relative to consumer accounts, but the educational finding remains open until tool use is connected to protected student capacity.
Recommended investigation
Compare courses with access alone, access plus active faculty mediation, and protected non-AI work. Measure source accuracy, voice, delayed performance, faculty contact, workload, and use of alternatives.
Principal official sources [10] [11] [12] [13]
California Institute of Technology
Caltech connects institutional security guidance with instructor authority, writing pedagogy, explicit disclosure, ethical inquiry, and concern about the effect of AI on trust and assessment.
Evidence tableView 5 recordsHide 5 records
Evidence area | Published record | Current judgment |
|---|---|---|
Educational purpose | Teaching resources ask instructors to begin with learning and decide whether AI promotes or impedes it. | The principle is strong; common institute-wide outcome measures are unresolved. |
AI rule | Humanities and Social Sciences allows only uses explicitly authorized in course materials. Other courses retain instructor authority. | The operating boundary varies by division and course. |
Tools and privacy | Institutional guidance restricts disclosure of sensitive information to consumer tools and directs users toward responsible accounts. | Approved tools settings retention and student alternatives require current documentation. |
Integrity and authorship | Caltech examples use disclosure memos and warn against relying on inaccurate detectors. | Actual investigation procedures appeal outcomes and cross-division consistency are not public. |
Independent performance | Proctored exams writing instruction and discipline-specific assessment remain available. | Comparable delayed transfer across writing mathematics science and coding has not been publicly demonstrated. |
Preliminary finding
Caltech presents one of the strongest public connections among learning purpose, instructor authority, privacy, authorship, and trust. The evidence does not yet show that those protections produce durable transfer across disciplines.
Recommended investigation
Review representative science, mathematics, coding, and writing courses under the same framework. Compare AI-supported work with proctored explanation, error correction, oral defense, and delayed novel problems.
Principal official sources [14] [15] [16] [17]
Loyola Marymount University
Loyola Marymount University combines whole-person educational language with practical teaching guidance, protected enterprise tools, academic-honesty resources, and explicit attention to AI-free and AI-integrated course choices.
Evidence tableView 5 recordsHide 5 records
Evidence area | Published record | Current judgment |
|---|---|---|
Educational purpose | University materials connect academics with whole-person education and ethical responsibility. | The reviewed AI materials do not establish shared measures of judgment vocation or independent thought. |
AI rule | Faculty and departments set course use within academic-honesty privacy and security requirements. | The rule experienced by students must be verified course by course. |
Tools and privacy | The university provides protected tools and asks educators to consider access cost privacy accessibility and alternatives. | The public record does not establish how often alternatives are used or whether they are equivalent. |
Teaching practice | Resources describe AI-free AI-integrated and co-constructed policy approaches. | Availability of options does not establish implementation quality. |
Independent performance | AI-free anchor work can preserve authorship and baseline capacity. | No common oral or supervised verification standard was located. |
Preliminary finding
Loyola Marymount offers a promising agency-and-alternatives model. Its formation claim will become stronger when course choices are linked to protected demonstrations and comparable results.
Recommended investigation
Document one course from each policy model. Compare rule clarity, student choice, disclosure, privacy, protected performance, instructor workload, and learning after assistance is removed.
Principal official sources [18] [19] [20]
Thomas Aquinas College
Thomas Aquinas College organizes its common curriculum around original texts, tutorials, seminars, mathematical demonstrations, laboratory work, composition, sustained conversation, and public defense of conclusions.
Evidence tableView 5 recordsHide 5 records
Evidence area | Published record | Current judgment |
|---|---|---|
Educational purpose | The published curriculum joins intellectual freedom truth moral inquiry and a common sequence of direct study. | The formation architecture is clear; comparable outcome evidence is limited. |
AI rule | The California Code of Conduct restricts AI and AI-enabled search in the generation of academic papers including research outlining argument drafting and editing. | The scope is unusually clear. Exceptions accommodations enforcement and revision history need direct evidence. |
Attention and relationship | Device and internet restrictions preserve face-to-face study conversation worship recreation and residential community. | Published purpose and structure do not by themselves establish wellness effects. |
Independent performance | Tutorials seminars demonstrations laboratories compositions and the senior thesis require visible human performance. | Rubrics inter-rater reliability and comparable scoring are not fully public. |
AI competence | The AI and Human Flourishing initiative studies technology through philosophical moral and civic questions. | It remains unclear how students develop practical responsible-use competence for later professional life. |
Preliminary finding
The college shows the strongest alignment between a restrictive AI rule and a preexisting formation-centered curriculum. It should not be declared superior on learning or wellness without comparable outcomes and accessibility evidence.
Recommended investigation
Preserve the full chain for composition and senior-thesis work, including drafts, tutor feedback, oral defense, accommodations, appeals, student experience, and later ability to use AI responsibly without surrendering judgment.
Principal official sources [21] [22] [23] [24]
K through 12 findings
The K through 12 track gives greater weight to developmental sequence, parent authority, consent, family capacity, foundational knowledge, protected relationships, and age-appropriate independence. A policy suitable for an advanced high-school seminar may be unsuitable for an elementary language or number lesson.
Evidence tableView 3 recordsHide 3 records
Institution | Most established strength | Principal unresolved question |
|---|---|---|
Mother of Divine Grace School | Parent-mediated classical stages and transparent formation purpose | Age-specific AI rules approved tools privacy and consent |
Harvard-Westlake School | Integrity independent inquiry wellness support and an AI ethics course | The operative schoolwide permissions disclosure privacy and assessment rules |
High Tech High | Project-based learning public exhibition student voice and board adoption of an AI policy | The full adopted policy and individual mastery within collaborative projects |
Developmental judgment
The published record does not justify broad student-facing access across the track. Each school should identify what knowledge and self-command must be established before AI assistance, which uses require parent or guardian involvement, what data may be collected, and how the student will demonstrate authorship and understanding without generation.
Family and wellness context
COVID-era disruption remains an interpretation condition. Students who entered disruption with weak attention, unstable family support, limited friendship, anxiety, or prior learning difficulty may have experienced a different trajectory from students with strong relationships and routines. Institution-level AI claims should not attribute later outcomes to a tool without accounting for those prior conditions and subsequent recovery.
Mother of Divine Grace School
Mother of Divine Grace School places primary instructional responsibility in the family and organizes learning through developmental stages that emphasize memory, imagination, language, logic, composition, discussion, and judgment.
Evidence tableView 5 recordsHide 5 records
Evidence area | Published record | Current judgment |
|---|---|---|
Educational purpose | The statement of purpose and methodology describe a coherent classical and Catholic account of intellectual and moral formation. | The public architecture is strong; comparable outcome evidence remains limited. |
Family role | Parents direct daily education with program materials consultation and online support. | Actual family burden variation support quality and student experience require participant evidence. |
AI rule | No complete public generative AI policy approved-tool list age sequence or consent framework was located in the reviewed record. | The correct status is unresolved rather than presumed absent. |
Independent performance | Narration recitation handwriting composition logic and discussion protect direct mental work. | Comparable scoring delayed transfer and the effect of online tools are not publicly established. |
Wellness and COVID | The school distinguished prepared home education from emergency crisis schooling and recognized isolation fear disrupted family life and lost friendships. | The account supplies context but not a causal comparative wellness study. |
Preliminary finding
The school supplies the clearest family-mediated developmental model in the sample. Its central AI-era need is a published age-specific governance framework that preserves the method while addressing tools privacy consent and later responsible competence.
Recommended investigation
Develop and evaluate stage-specific rules with parent participation. Begin with protected composition and oral explanation, then compare any bounded AI use only after foundational knowledge and authorship are established.
Principal official sources [25] [26] [27] [28]
Harvard-Westlake School
Harvard-Westlake publishes a broad college-preparatory curriculum, independent research opportunities, health and wellness support, and a course examining ethics technology and society in the age of artificial intelligence.
Evidence tableView 5 recordsHide 5 records
Evidence area | Published record | Current judgment |
|---|---|---|
Educational purpose | The school emphasizes educational excellence integrity character community curiosity and purpose beyond the self. | These aims are clear but not yet connected publicly to a common AI-era performance standard. |
AI curriculum | The current course guide includes structured inquiry into artificial intelligence and society. | A course description does not establish the full syllabus assignments student work or learning results. |
AI operating rule | A complete public framework for permissions disclosure approved tools privacy age consent and discipline was not located. | Rules may exist behind authenticated portals; the public status remains unresolved. |
Independent performance | Writing research arts and other conventional course work provide opportunities for human authorship. | No public recurring oral defense supervised writing or source-audit requirement was established. |
Wellness and relationships | The school publishes counseling wellness and relationship-centered commitments. | No public longitudinal evidence connects AI use with sleep attention friendship belonging or help seeking. |
Preliminary finding
Harvard-Westlake has the curricular and support capacity to study AI as an ethical and human question. The public record is not yet sufficient to judge how student use is governed across grades and courses.
Recommended investigation
Preserve the operating policies and one full AI-ethics inquiry. Compare student-authored writing, oral defense, source verification, consent, privacy, workload, and wellness by grade and prior support condition.
Principal official sources [29] [30] [31] [32]
High Tech High
High Tech High provides a distinctive project-based case in which inquiry, making, critique, collaboration, adult-world connection, and public exhibition can make student work authentic and visible.
Evidence tableView 5 recordsHide 5 records
Evidence area | Published record | Current judgment |
|---|---|---|
Educational purpose | The design principles emphasize equity personalization authentic work and collaborative design. | The principles support formation but do not specify the individual mental capacities AI must not displace. |
Policy governance | Board minutes show approval of an Artificial Intelligence Use Policy on May 28 2025 after staff and student involvement. | The full final policy text was not located in the public record reviewed. |
Student participation | A December 2023 student presentation recommended protecting grammar foundations and using AI as a guide rather than an answer. | The presentation documents deliberation rather than the operative rule. |
Independent performance | Projects critique and exhibition provide authentic evidence of participation and communication. | Group products can conceal individual mastery unless paired with explanation or supervised construction. |
Outcomes and wellness | Public accountability and project evidence are available across the network. | No public evaluation links the AI policy to mastery equity attention belonging family burden or adult support. |
Preliminary finding
High Tech High shows the strongest published combination of student voice and formal board action in the K through 12 group. A complete substantive judgment requires the adopted policy and evidence that every student can explain and reproduce the intellectual contribution credited to that student.
Recommended investigation
Publish or preserve the final policy and review a representative project from elementary through high school. Add individual oral defense, source verification, protected construction, parent notice, privacy review, and delayed transfer.
Principal official sources [33] [34] [35] [36]
Findings across the ten institutions
Formation and governance do not yet coincide
The institutions with the clearest formation purposes do not necessarily publish the most complete AI governance. The institutions with the most mature infrastructure do not necessarily define the human capacities their systems are intended to preserve. The project should retain separate judgments for formation, governance, implementation, independent performance, and outcomes rather than compressing them into one score.
Course authority remains necessary
A uniform campus rule cannot decide every educational use. The same tool can supply an accessibility aid, a useful critique, or the answer that displaces the lesson. Faculty and teachers need authority to protect the mental act, while institutions remain responsible for privacy, accessibility, consistency, records, safety, and fair process.
Approved tools are an incomplete measure
Managed accounts can improve contractual control and reduce exposure compared with consumer accounts. They do not prove that the activity is educationally appropriate, that students understand the data flow, that an alternative is meaningful, or that learning survives tool removal.
Assessment must become more visible
The strongest institutional response is not detector-based policing. It is recurring protected work, process evidence, oral explanation, source verification, delayed transfer, and fair review. Publicly reported examples exist, but no institution in the group has yet demonstrated this complete architecture across its programs.
Wellness evidence remains especially weak
The public record contains wellness services, institutional commitments, COVID accounts, and warnings about dependence or human support. It does not yet show the long-term effect of educational AI on attention, sleep, friendship, belonging, anxiety, self-command, help seeking, or family relationships.
No final ranking
A final ranking would imply comparable evidence that does not exist. The defensible output is a set of preliminary findings with explicit evidence classifications and institution-specific verification priorities.
California institutional source key
These 36 official institutional sources resolve the numbered references used in the ten California case findings. They document published policies, guidance, programs and institutional positions; they do not by themselves prove implementation or student outcomes.
- UC Berkeley AI HubCentral directory of campus tools training policies and opportunities.
- UC Berkeley teaching guidance for AITeaching and course-design resources.
- Berkeley Law artificial intelligence policy questions and answersUnit policy effective summer 2026 restricting AI assistance in credited work except where authorized.
- UCLA AI in Teaching and LearningSix principles literacy module Faculty Fellows workshops and advisory structure.
- UCLA guiding principles for responsible useInstitutional responsibility risk and data guidance.
- UCLA AI use and recommendation guideApproved enterprise tools and use guidance.
- Stanford AI and Your LearningStudent guidance on permissions disclosure learning accuracy bias sources and human support.
- Stanford course policy guidanceInstructor guidance for aligning course rules with learning goals.
- Stanford COLLEGE 102 student policy project2026 account of more than 1200 students developing AI-era assessment proposals.
- USC faculty guidance for the ChatGPT Edu rolloutJanuary 2026 guidance preserving faculty course authority.
- USC ChatGPT EduManaged university account information and support.
- USC policies and policy governanceOfficial policy home including the Artificial Intelligence General Policy updated in 2026.
- USC academic integrity guidance for generative AIStudent and faculty guidance connecting course rules and academic responsibility.
- Caltech AI in TeachingCentral teaching resource and policy route.
- Caltech Humanities and Social Sciences policy on generative AIExplicit authorization rule tied to course learning.
- Caltech resources for teaching in the age of AICourse-policy assessment and misinformation guidance.
- Caltech generative AI security guidanceInstitutional guidance on protected information accuracy and responsibility.
- Loyola Marymount AI in Teaching and LearningTeaching options policy examples and educator support.
- Loyola Marymount AI tools and servicesInstitutional tools access and use information.
- Loyola Marymount academic honesty and AIAcademic-honesty guidance and course responsibility.
- Thomas Aquinas College California Code of ConductCurrent academic and technology rules.
- Thomas Aquinas College syllabusCommon curriculum tutorials seminars laboratories composition and senior thesis.
- Thomas Aquinas College AI and Human FlourishingPublic initiative examining artificial intelligence and human flourishing.
- Thomas Aquinas College online transition accountInstitutional account of emergency remote instruction and engagement limits.
- Mother of Divine Grace School statement of purposeEducational mission and account of formation.
- Mother of Divine Grace curriculum methodologyDevelopmental method and family-centered instruction.
- Mother of Divine Grace curriculum stagesPrimary grammatical dialectical and rhetorical sequence.
- Mother of Divine Grace Crisis Schooling Is Not HomeschoolingInstitutional account distinguishing prepared home education from COVID crisis schooling.
- Harvard-Westlake course descriptionsCurrent curriculum and artificial-intelligence ethics course description.
- Harvard-Westlake Kutler CenterIndependent interdisciplinary research and inquiry resources.
- Harvard-Westlake health and wellnessPublished student support and wellness resources.
- Harvard-Westlake mission and school overviewInstitutional language concerning excellence integrity character community and purpose.
- High Tech High May 2025 board minutesBoard approval record for the Artificial Intelligence Use Policy.
- High Tech High December 2023 board minutesStudent policy recommendations and board discussion.
- High Tech High design principlesEquity personalization authentic work and collaborative design.
- High Tech High student and parent handbook routePublic route for current school handbooks and operating rules.