Joint Senate-Administration Committee Report on AI Use in Academic Contexts

August 27, 2026

In Fall 2025, we jointly convened an Academic Senate-Administration Advisory Committee on the Uses of Artificial Intelligence in Academic Contexts to develop guidelines and guardrails for the responsible use of AI in teaching, research, and academic administration. The committee (co-chaired by Lisa Parks, Professor of Film and Media Studies, and Timothy Sherwood, Professor of Computer Science and Dean of College of Creative Studies, with membership drawn from faculty, staff, and administration) met from December 2025 through June 2026, consulting broadly with stakeholders and reviewing precedents across the UC system.

Their final report, which we transmit with this message, recommends a framework for course-level AI policy and offers further recommendations on academic integrity, personnel review, campus access, and campus strategy. The report represents committee consensus on core issues and notes others about which discussion and debate continues.

From it we extract five principles:

Academic Oversight. While technical implementation involves IT and administrative offices, questions of pedagogy, scholarship, evaluation, and intellectual integrity should be overseen by the Senate and Academic Affairs.

Critical Literacy. A primary goal is critical AI literacy and informed, ethical engagement with these technologies. Equitable access should serve this goal rather than substitute for it.

Data Sovereignty. Student and faculty intellectual output must not become proprietary training material. The campus should set the terms governing the use of academic work rather than accept a vendor’s defaults. This position aligns with recommendations from the UC Academic Senate Workgroup on Artificial Intelligence.

Academic Integrity. Automated AI detection tools should not serve as the sole basis for academic integrity proceedings. To preserve integrity, instructors are encouraged to consider modes of assessment such as drafts and revision, oral demonstration, or other work done by students in the room.

Institutional Control. The campus should not rely on a single AI vendor. Models and the companies behind them are changing quickly, and institutional control over academic decisions and data depends on our not being exclusively tied to any single one.

Consistent with the report’s recommendation to pursue a multi-model approach, the campus is now providing faculty and staff access to an AI Commons. This university-managed platform offers a secure interface to multiple model families—including Claude, Qwen, Llama, and others, with more to come—and ensures, by contract, that vendors cannot train on or retain academic and research material, including unpublished work. Faculty and staff also continue to have access to Gemini through the campus’s Google Workspace environment. As the report also recommends, we expect that subsequent access for students will be accompanied by guidance about academic integrity.

We appreciate multiple efforts by ITS, the Office of Teaching and Learning, the Library, and the AI Community of Practice to make resources and guidance available; ongoing collaborative work will unify that guidance and keep it responsive as the technology evolves. The AI tools that many of us have been using answer a question or draft text when prompted, but the line between “using a computer” and “using AI” is already blurring in search engines and word processors. A newer generation goes further and is built to act on a user’s behalf: browsing and completing tasks with minimal further instruction. Agentic systems, along with research strategy and admissions tools, were among the questions one committee could not address in a year. These and other topics will need attention on campus and in the systemwide conversations already underway.

We are grateful to the advisory committee for their demanding, careful work. Going forward, we plan to establish a standing committee, with representation from the Senate and the administration, to develop policies and modes of AI governance that reflect the mission of our public research university.

Thank you.
 

July 31, 2026

The rapid and accelerating use of AI and widespread access to a variety of AI tools call for the coordinated development of campus guidelines, policies, and principles for the use of AI in different academic contexts. A recent survey of University of California faculty indicated that a significant number use or intend to use AI tools for teaching, student engagement, and research, and see potential benefits in these aspects of their work. However, faculty also identify challenges and concerns related to impacts on student learning, critical thinking, and originality; data privacy and security; bias and equity; intellectual property; academic integrity and the difficulty of detecting AI-generated content; and errors and misinformation. Students, staff, and faculty, through various forums, have expressed a need for greater shared understanding and clarity in these and other areas related to AI in the university.

In this context, Executive Vice Chancellor/Provost David Marshall and Academic Senate Chair Rita Raley jointly convened the AI Advisory Committee in Fall 2025. The committee's charge was as follows:

This joint Senate-administrative committee will help the campus develop both guidelines and guardrails for the responsible use of AI in teaching, research, and academic administrative activities. General guidance and recommendations about principles, protocols, and best practices will help colleges, schools, departments, and disciplines develop more specific guidelines that reflect their needs and concerns.

Additional topics for review include the use of AI in evaluating and assessing faculty, students, and staff; governance; cybersecurity; coordination with the development of workplace AI protocols and practices; pedagogical innovation; and guidance about potential AI tool implementations.

While the committee does not have decision-making authority nor does it speak for the Academic Senate or the university administration, it was able to consult broadly with stakeholders across campus, identify areas where cross-cutting guidance is needed, and offer recommendations to the appropriate campus bodies (e.g. the Academic Senate, the EVCP's office, the Chancellor's Office, department chairs, and others) who can then act on through their own processes. The recommendations in this report are offered in that spirit.

The committee is co-chaired by Professor Lisa Parks (Film and Media Studies) and Dean Timothy Sherwood (College of Creative Studies) and reports jointly to EVCP Marshall and Senate Chair Raley, who serves ex officio. Committee members were drawn from across the faculty, staff, and administration to represent a wide range of disciplinary perspectives and institutional functions. Members include Associate Vice Chancellor Linda Adler-Kassner, Professor Tarek Azzam, CIO Josh Bright, ITS Director Bret Brinkman, Professor Miguel Eckstein, Assistant Vice Chancellor James Ford, University Librarian Todd Grappone, Professor Renee Houston, Associate Dean Chandra Krintz, Professor Christopher Kruegel, Research Integrity Director Barry Rowan, Deputy CIO Joe Sabado, and Professor Xifeng Yan.

The committee held its first meeting in December 2025 and met monthly through June 2026, for a total of seven one-hour meetings. The first several meetings constituted an extended intake phase. During this period, the committee reviewed the UCSB Senate Working Group’s Preliminary Recommendations for Use of Artificial Intelligence Tools (July 2025) and the UC System-wide Academic Senate AI Workgroup Report, received briefings from CITRAL and OTL on the current teaching and learning landscape, heard from Dean of Students Joaquin Becerra on how student conduct and academic integrity processes work at UCSB, met with AVC Dana Mastro and CAP Chair Laury Oaks on the intersection of AI and academic personnel review, and discussed the campus AI ecosystem with stakeholders from IT, the Library, and various academic units. The committee also met with John Longbrake and Alex Parraga to discuss AI and campus public relations and consulted with the Graduate Division and the Academic Senate’s Graduate Council. Finally, we took note of conversations already underway across the UC system, where several campuses have been grappling with similar questions.

Because AI is genuinely cross-cutting, the committee had to make deliberate choices about where to focus. Not only do decisions made in one traditional area of governance often have impacts that span many others, but we also could not responsibly address every dimension of AI at the university in a single year. To prioritize, the committee came up with a list of priorities related to AI on campus and then voted on which areas to develop recommendations for. The topics considered included AI and teaching and learning; AI and academic personnel policy; AI and the campus's public profile; AI governance; AI in administrative functions; AI in academic programs and research strategy; and AI in admissions. The committee organized its work in the three general areas that received the strongest support:

• Teaching and Learning: the most immediate student-facing need. Students currently navigate fully bespoke and often contradictory AI policies across their courses, with little shared language or logic. This area also encompasses questions of student access, equity, and how the campus supports faculty in developing effective AI policies.
• Academic Personnel: AI is increasingly present in the faculty merit and promotion process, from letter writing to preparation of candidate materials, and departments have limited guidance on how to approach it.
• Public Profile and Governance: There is currently no point of coordination for AI activity at UCSB, which means no coherent strategy and no way to represent what the campus is doing in a unified way.

From these three priority areas, the committee developed five recommendations, which are attached to this memo:

1. A Shared Course Policy Framework (Green / Yellow / Red). A three-tiered framework for communicating AI use expectations at the course level, consistent with the UC System-wide Senate recommendation for a "center-periphery approach." Green permits and may encourage AI use; Yellow permits it under instructor-set conditions; Red does not permit it. The framework provides a shared vocabulary while preserving full instructor autonomy.
2. AI and Academic Integrity. AI detection tools should not be used as the sole or primary basis for academic integrity proceedings. The committee recommends investment in assessment design that is resilient to AI misuse by design and in educational rather than punitive responses when appropriate.
3. Campus AI Access. The committee analyzed the current landscape of student AI access, evaluated institutional options, and presented its analysis and recommendation. While there are myriad real concerns, access is both an equity issue and a data privacy issue.
4. Use of AI in Academic Personnel Review. Drawing on a dedicated session with AVC Dana Mastro and CAP Chair Oaks, the committee addresses AI's role in the faculty review process from administrative support to evaluative judgment, and from candidate-controlled to university-controlled uses.
5. Campus AI Strategy: AI Governance, Coordination, and Public Profile. The committee recommends that the campus assign clear leadership and coordination responsibility around AI, integrate AI into the upcoming institutional strategic planning process, and build a coordinated public profile from that coordination work.

The committee's goal throughout has been to capture the consensus view of its members where it exists, and to be transparent about where members hold different perspectives. These recommendations reflect genuine agreement on core principles, even where committee members have different views on specific details. Where those differences exist, we have tried to present those diverging perspectives honestly and clearly.

These recommendations are directed to a range of campus bodies, and the committee recognizes that implementation will require collaboration across institutional lines. We hope that they provide a useful starting framework and that the conversations they prompt will be as productive as the ones that produced them.

Finally, the committee wants to be clear about what it did not address. The topics that were not prioritized in our vote (administrative AI use, AI in academic programs and research, and AI in admissions) are not unimportant. They simply require their own focused attention. The committee encourages the campus to take them up, whether through a continuation of this body or through other mechanisms. UC system-wide conversations on AI policy are also actively underway, and several of our recommendations intersect with that work. We encourage the campus to stay closely connected to those system-wide efforts as implementation proceeds.
 

COMMITTEE RECOMMENDATIONS

1.    A Shared Course Policy Framework (Green / Yellow / Red)

Summary of Recommendation
Collaboratively establish a university-wide, three-pronged course policy framework, with default guidance that applies when instructors have not specified a course policy of their own.

Directed To: Academic Senate (Undergraduate Council, Graduate Council); Office of Student Conduct; Center for Innovative Teaching, Research, and Learning. 

Context
Students currently face fully bespoke and often contradictory AI policies across their courses, with little shared language or logic. One instructor may encourage AI use, another may prohibit it, and a third may be completely silent on the topic, even all within the same department. Faculty, meanwhile, support developing, communicating, and enacting their policies effectively. The result is stressful confusion for students and inconsistency that risks undermining trust in teaching, learning, and academic integrity processes. This framework allows faculty to have autonomy in how they use AI in their teaching.

Details of Recommendation
The committee recommends that the university work collaboratively across academic senate and administrative lines to develop and adopt a campus-wide three-pronged framework (Green / Yellow / Red) for communicating AI use expectations at the course and assignment level:

Green -- A first policy under which AI use is generally permitted and may be encouraged for this work. Students may use AI tools freely as part of their process, and the policy should spell out both expectations and responsibilities under such a model.

Yellow -- A second policy under which AI use is permitted under specific conditions set by the instructor. Conditions may include disclosure of use, attribution of AI-generated content, restrictions to specific stages of work (e.g., editing but not initial draft), or use of specific tools only. Such a policy should allow faculty to easily and clearly extend the policy in ways that are unique to their disciplinary and pedagogical needs.

Red -- A third policy under which AI use is not permitted. This is to support an instructor who has determined that the learning objectives require unassisted student effort. Exactly what that means and what the ramifications of violation are should be clearly spelled out by the instructor.

The committee recommends the development of a policy under which every course syllabus includes a clear statement about which of these three options will be followed in the class, noting any refinements (e.g. specific additions or restrictions) where necessary. The committee also understands that some instructors will choose to leave AI policy in their class unaddressed. As such, the committee recommends that one of the three be selected as the default policy in the absence of other guidance.

Key Considerations
The UC system-wide Academic Senate AI Workgroup Report recommends a “center-periphery approach”: the campus provides a central framework flexible enough to accommodate instructor and discipline-specific practices ranging from “active incorporation of AI tools to their outright prohibition, depending on pedagogical goals and disciplinary norms.”

A consistent framework makes academic integrity processes more defensible: if expectations are clearly communicated in standardized language, violations are easier to identify and adjudicate.

This approach preserves instructor autonomy over pedagogical choices while significantly improving the student experience of navigating different AI policies across courses. Student input should be solicited before this policy framework is finalized, and instructors should be encouraged to contextualize their approach within the learning goals of the course.

Complicating Factors
The Senate has the authority, should it choose to exercise it, to require a syllabus statement regarding AI. Meaningful uptake, however, depends on designing a framework that is genuinely useful, not burdensome.

The “Yellow” option carries the most ambiguity as “permitted with conditions” can mean so many different things across disciplines. Without good examples and guidance, Yellow becomes a catch-all that recreates the current confusion. The default options (e.g., well-defined checklists) must cover a broad set of options, but not be so voluminous that they cannot be understood easily in their entirety.

AI tools are increasingly embedded in standard software (word processors, search engines). The line between “using AI” and “using a computer” is being purposefully blurred, meaning the framework must be revisited regularly. Students with disabilities may rely on AI-powered assistive technologies.
Restrictive policies need explicit accommodation language to avoid conflict with accessibility requirements.

2.    AI and Academic Integrity

Summary of Recommendation
Violating conventions of academic integrity through inappropriate AI use is a real concern, but detection tools should not serve as the basis for initiating academic integrity proceedings. A detection score might contribute to an instructor's suspicion, but it should not constitute more than corroborating evidence, which alone is insufficient to sustain a charge. The committee recommends that the university promote assessment practices that are resilient to AI misuse by design while ensuring faculty, students, and staff have a clear and shared understanding of responsibilities around AI use in the classroom. There should also be clear paths of action when those responsibilities are not met.

Directed to: Office of Student Conduct; Academic Senate (Undergraduate Council, Graduate Council); Center for Innovative Teaching, Research, and Learning; Department Chairs and Program Directors.

Context
Some instructors have turned to automated “AI detection” tools as a primary means of identifying violations of academic integrity, a concept that is put into practice differently across disciplines and contexts (Bretag, 2016). These tools claim to distinguish between human-written and AI-generated text by analyzing language patterns. Based on extensive published research, the committee finds that many of these tools are demonstrably unreliable and that their use as the basis for integrity charges poses serious risks to students and to institutional credibility. The evidence against the reliability of current AI detection tools is substantial and growing. Despite this, there is increasing use of models such as Pangram, described as the “most reliable and most accurate AI detection tool in the market by third parties, including the University of Maryland and the University of Chicago.”

Accuracy collapses under real conditions. The technology here is moving so quickly it is hard to have confidence in any of the reports, especially if individuals are engaged in counter-efforts. For example, one study found that while detectors identified AI-generated text with 74% accuracy on clean samples, accuracy dropped to 42% when students made even minor edits. With some degree of prompt engineering, researchers are able to reduce Turnitin’s detection rate from 100% to 0%. (Perkins et al., published in academic integrity research, 2024-2025). As such, efforts are perhaps better expended in other ways.

Non-native English speakers are disproportionately harmed. Stanford researchers found that detectors misclassified over 61% of essays by non-native English speakers as AI-generated, while performing “near perfectly” on essays by native speakers. ESL students face 2-3x higher false positive rates (Liang et al.
2023), and when those students are questioned about AI use, many are not equipped to advocate for themselves.

In summary, detection tools work by identifying statistical regularities in AI-generated text. As models improve and their output becomes more varied and human-like, and as students learn to edit and adapt AI output, the statistical signal these tools rely on will only weaken.

Details of Recommendation
The campus should consider the adoption of a policy that 1) reinforces the importance of teaching conventions of academic integrity; and 2) makes clear that AI detection tools cannot serve as the basis for decisions or judgments based on perceived violations of academic integrity. If detection tools are used at all, they should serve only as one signal among many and should prompt a conversation with the student rather than a formal charge or immediate grade change. The burden of evidence in integrity proceedings must rest on more than an automated probability score, and students need an opportunity to contest an allegation of AI use when challenged. Faculty are encouraged to work with the Office of Student Conduct on these issues.

Invest in assessment design that is resilient to AI misuse. The most effective response to AI in academic work is not detection after the fact but assessment that makes misuse unproductive or visible by design. Approaches include (but are not limited to) process-based assignments that require students to show development of learning via iterative work (drafts, revision history, or annotations), in-class or oral demonstrations of competency, portfolio-based assessment that builds a body of work over time, and use of the university testing center as a controlled AI-free environment. We recognize that assessment redesign requires significant faculty time and institutional investment and must be balanced with other priorities.

Promote educational responses to suspected violations of academic integrity. Instructors need clearer guidance on the paths of action available when they suspect a student of using AI to willfully violate clearly defined practices of academic integrity, so they may escalate appropriately and constructively. In the classroom, this work would be best facilitated by clear policies around AI use (e.g. as outlined under the red/yellow/green policies). Where a student has clearly and deliberately misrepresented AI-generated work as their own in violation of a stated course policy, existing conduct processes apply. Faculty need to be very aware of their rights and responsibilities in this process, and the university should make a concerted effort to ensure all parties are informed about the stakes of violating academic integrity.

Key Considerations
Instructors retain full authority over how they evaluate student work. This recommendation does not limit an instructor’s ability to design assignments, set expectations, or assess student learning. It affirms that authority by encouraging instructors to design assessments that genuinely measure what they intend to measure, rather than relying on a third-party tool to police compliance after the fact.

A clear course policy is an essential complement. Clear communication of AI expectations at the course level reduces the ambiguity that leads to both AI misuse and false accusations. When students know what is expected, and when those expectations are communicated in shared language, both students and instructors are on firmer ground.

Students with disabilities may rely on AI-powered assistive technologies. Any guidance on AI use in assessment must include explicit accommodation language to ensure that accessibility tools are not inadvertently flagged or restricted. Coordination with the Disabled Students Program is essential.

Complicating Factors
Some instructors have already integrated detection tools into their workflow and may view this recommendation as undermining their ability to maintain academic integrity. The committee acknowledges this concern and emphasizes that the recommendation is not to ignore potential misuse but to pursue it through more reliable means. Faculty development and support in resilient assessment design will be important. CITRAL is an excellent resource with relevant expertise and is working on a programmatic assessment pilot project.

The line between “using AI” and “using standard software tools” continues to blur. Spell-checkers, grammar tools, and search engines all incorporate AI to varying degrees. Any framework for evaluating AI use must grapple with the fact that the boundary is not clean and will only become less so.

Students are aware of the limitations of detection tools. A policy that relies heavily on detection may create an adversarial dynamic in which students optimize for evading detection rather than engaging with the learning goals of assignments. Assessment design that makes the learning visible is both more effective and more conducive to the educational relationship, and as technology continues to develop, we will likely need to continue to reconsider how academic competency is evaluated and certified.

3.    Campus AI Access

Summary of Recommendation
The committee recommends adopting a multimodal approach to campus AI access by providing a diverse range of AI tools to faculty, staff, and students, accompanied by no-cost access, training, and resources that promote critical AI literacy and responsible use. It also recommends that AI procurement decisions follow a consultative process to ensure campus AI investments reflect the campus’s academic mission and remain responsive to the rapidly evolving AI landscape.

Directed to
Provost/EVC; Chair, Academic Senate; AVC for IT

Context
The landscape of LLM use is changing rapidly, with several commercial providers competing for university contracts. UCSB has held a contract with Google for years and is piloting Gemini among faculty and staff in 2025/26. Under current terms, UCSB has the option to make Gemini available to all UCSB community members and has also licensed other models (Llama, Claude, Amazon, Qwen 3) through a token system. Against this backdrop, the committee considered whether UCSB should make some AI tools accessible to all and, if so, which.

The committee’s deliberations were extended and substantive, reflecting a diversity of perspectives rather than a settled consensus. Members worked first to reframe the question. Several cautioned against framing the decision as a binary choice between vendors—Google or Anthropic—and urged the committee to begin with outcomes: what does it hope to achieve? The view that gained the most traction was that the primary goal should be cultivating critical AI literacy across the UCSB community, with equity of access as a companion concern. This reframing anchors a fast-moving, uncertain topic in institutional purpose rather than in the features of any one product.

The discussion was attentive to the broader landscape. Members noted that every UC campus except Santa Cruz now provides some form of AI access, that some faculty already require students to use AI tools, and that ChatGPT has been in widespread use for over a year. Some characterized the resulting pressure as a kind of institutional “fear of missing out,” but the committee was careful to distinguish reactive imitation from evidence-based learning. Beyond the UC, members pointed to the concerning OpenAI contract with the Cal State system and the problems that have followed. There was candid acknowledgement that implementation has often outpaced consultation, that decisions sometimes had to be walked back, and that adequate deliberation before commitment is itself a value worth protecting.
Members thus agreed that any recommendation should reach an extensive set of addressees and advance through dialogue with the provost, the senate chair, and others–as part of a considered process rather than a fait accompli.

A unifying theme was the volatility of the field. Models, and the corporate circumstances behind them, are changing rapidly; the recent disruption surrounding Qwen 3.5 was cited as a vivid illustration. Several members expressed strong reluctance to commit the campus to a single model or vendor, favoring investment in infrastructure that equips UCSB community members with an understanding of various AI tools. Members also cautioned that focusing on conversational models alone may be too narrow, since agentic AI systems, which can sign in as students and act on their behalf, raise distinct and pressing questions, including for assessment.

Details of Recommendation
A multimodal AI approach. The committee recommends adopting a multimodal approach to campus AI access by making a diverse range of AI tools available to faculty, staff, and students. Committee members expressed significant reservations about committing to a single AI model or vendor, citing the rapid pace of technological change and the risks of vendor lock-in. A multi-modal strategy provides greater flexibility, supports a wider range of teaching, research, and administrative needs, and allows the campus to adapt as the AI landscape evolves.

Academic framing of AI access. The committee recommends framing campus AI access as an academic initiative centered on developing critical AI literacy. Advancing AI literacy requires making AI tools broadly available to faculty, staff, and students at no cost to individual users, accompanied by training and support that enable informed, ethical, and effective use. Existing campus resources, including CITRAL, the Library’s Carpentry workshops, and relevant research and admin units, can play a key role in providing this training and fostering a campus-wide culture of responsible AI use.

Consultative AI procurement. The committee recommends that decisions regarding the procurement of AI tools for broad campus access follow a transparent and consultative process that includes both academic and administrative stakeholders. Procurement decisions should be guided by the university’s core missions of research, teaching, and public service, while remaining responsive to the evolving AI ecosystem and the diverse needs of the campus community.

Key Considerations
If the central aim of access is critical AI literacy, that literacy is best built through structured engagement rather than left to chance. This makes broad, supported access the most effective instrument for that goal. The pursuit of digital equity reinforces this case: because many students already rely on AI tools, disparities in access to paid services create unequal learning conditions by limiting some students to weaker models and more restrictive usage limits. Providing campus-supported, no-cost access to high-quality AI models helps ensure equitable access to educational resources and supports the university’s commitment to inclusive student success.

Because models and the firms behind them change rapidly, there is clear value in avoiding commitment to any single vendor. A multimodal approach built on infrastructure such as AWS lets the campus offer a range of token-based models—Claude, Llama, and Qwen among them—while controlling costs and preserving the ability to adapt.

Building on university-managed infrastructure rather than a consumer subscription keeps academic decisions and data under institutional control. A token-based approach lets the campus set the terms on which student work, research materials, and instructional content are handled, rather than accepting whatever a vendor’s default agreement provides. The committee’s dissatisfaction with the shifting storage terms of the prior Google contract is a concrete reminder of what is at stake when those terms are set elsewhere. For a public university, this alignment is not incidental: it ensures that decisions about pedagogy, privacy, and the handling of academic work remain within the institution answerable for them, rather than drifting to a vendor whose priorities may diverge from UCSB’s.

Complicating Factors
Three tensions complicate the path forward. First, access alone does not produce AI literacy. Even if UCSB makes AI tools available, there is no guarantee that they will be adopted and used by faculty, students, and staff. (This was a core concern with the Cal State system’s adoption of ChatGPT; hundreds of thousands of licenses went unused.) Without attention to training and incentives, even well-chosen tools may not advance the AI literacy that the recommendations are meant to serve. It would be helpful to consult adoption research when making decisions about procuring broadly accessible AI tools.

Second, the recommendations address only a portion of the actual landscape. Two important categories fall outside of their reach. First, agentic systems that can act on a user’s behalf raise distinct and pressing questions for assessment and academic integrity. They differ in kind from the LLMs and chat interfaces that the committee focused on. Second, closed models such as ChatGPT, which are already widely used, cannot be integrated with an infrastructure-based approach. While important aspects of the rapidly evolving AI landscape remain beyond the scope of these recommendations, inaction risks leaving the university behind as peer institutions continue to advance.

Finally, the committee identified the university’s relationship with Google and the governance of campus AI access decisions as important areas of ongoing debate. As noted, some committee members expressed concern about the shifting terms of the previous Google contract and worry that deeper institutional commitment could further commercialize the campus digital environment and increase the university’s dependence on a single corporate partner. Others regard Gemini as a pragmatic, cost-effective way to extend campus AI access now. Underlying these concerns is a broader unresolved governance question: whether AI access should be procured at the campus level or through the UC system.

4.    Use of AI in Academic Personnel Review

Summary of Recommendation
Before the next academic personnel (AP) review cycle, advance a shared, discipline-grounded approach to AI in AP review through four coordinated steps: faculty-led conversations; jointly issued campus guidance distinguishing administrative assistance from evaluative judgment; explicit protections for confidentiality of candidate data; and measures to protect the right to human evaluation.

Directed to
EVC/P; Academic Senate; AVC for Academic Personnel; Committee on Academic Personnel; Deans; Department Chairs & Managers

Context
Generative AI is now embedded in scholarly and administrative workflows, and its presence in AP review has outpaced shared norms. Awareness and practice are uneven across disciplines: some treat AI as a “dirty word,” others use it productively, and there is no clear picture of which departments are using it. Chairs are divided between near-prohibition and assumed widespread use. Departments are not required to disclose use of other writing tools in case letters, raising the question of whether AI tools should be treated differently. Candidates, meanwhile, may use AI to draft self-assessments and other dossier materials. While AI can serve as a time-saving aid, human experts must evaluate cases against disciplinary standards, and the core commitments to equity, consistency, and transparency must hold throughout. The absence of campus-wide guidance ahead of the upcoming AP review cycle is leaving questions about AI use unsettled, and no single unit should bear sole responsibility for establishing that guidance.

Details of Recommendation
Convene departments or units for disciplinary conversations in the coming academic year. Each department and/or research unit should hold a structured discussion, anchored in discipline-specific standards and informed by professional societies and journals, about how and when AI tools may be used in candidate materials, letters, and the evaluation of scholarship. Aim for shared expectations, rather than prescriptive rules; recognize that standards will need recalibration as AI shifts the balance between quality and quantity. In the near future, CAP/AP may need to determine how to credit scholarship that is aided to a non-trivial extent by LLMs, and also may need to revisit ARO guidance.

Establish coordinated governance through AVC AP within Senate parameters. AVC AP should coordinate with the Senate, CAP, and Dean’s offices to gather information on current AI practice and circulate brief guidance before the next AP cycle. That guidance should be framed around human evaluation, equity, confidentiality, and transparency.

Protect confidentiality and candidate data. Communicate clearly to the campus community that confidential personnel documents must be handled according to university policy and state and federal laws.

Protect the right to human evaluation. Insist that the AP review process be conducted by human experts, internally and among external reviewers. Emphasize the requirement of human evaluation in campus AP training workshops. Revise templates for external letter solicitations to signal both the requirement for human evaluation and the importance of protecting candidates’ data. For instance, include this language in the letter template: “To prevent violations of the author’s intellectual property (IP), confidentiality and data privacy, including personally identifiable information or AI developers claiming ownership of inputted content or expose sensitive data: Unpublished manuscripts (including full books, chapters and journal articles) must not be uploaded to any AI tools for any purposes.” Ask external letter writers to sign a certification confirming their letters were written by a human author and not AI.

Key Considerations
Departments set the terms and standards of evaluation, but the responsibility for AI in AP review is shared across the EVC/P, Senate, AVC AP, CAP, Deans’ offices, and departments, and any formal action will require Senate approval. Within that shared structure, the core commitments of equity, consistency, and transparency must be prioritized, given the potential for AI use to create inequities across cases, candidates, departments, or external letter writers. Human judgment remains at the center: AI may assist with administrative tasks, but the evaluation of scholarship and creative work, teaching, and service must remain a human function. AP deliberations and confidential personnel materials must be protected from third-party AI tools. The most durable path forward is structured conversation rather than intensive scrutiny or “policing” of behavior.

Complicating Factors
Several tensions complicate any guidance. Departmental autonomy over standards must be reconciled with CAP’s expectations of cross-case equity, and because departments are not asked to disclose other writing aids in case letters, any asymmetric treatment of AI tools must be approached carefully. Many faculty members have joint appointments, and departmental standards may conflict. External letters are a particular pressure point: AI authorship is difficult to verify; equity in external letter writing is already uneven; and suspicious external letters have already appeared. Beyond the review process itself, data retention and legal exposure raise broader concerns. Some third-party AI tools retain interactions indefinitely, implicating privacy, ownership, and individual rights questions that the Senate may need to address. Finally, the AI landscape is changing rapidly and varies sharply across disciplines; any guidance issued for the coming year should be revisited and updated rather than treated as settled.

5.    Campus AI Strategy

Summary of Recommendation
The committee recommends that the campus assign clear leadership and coordination responsibilities for AI to an individual or a designated body. There is currently no point of leadership or even coordination for AI activity at UCSB. Until that exists, the campus will struggle to develop a coherent strategy, a credible public profile, and effective cross-unit collaboration around this critical topic.

Directed To
Executive Vice Chancellor; Chancellor’s Office (in connection with upcoming institutional strategic planning); Office of the Chief Information Officer; Academic Senate.

Context
AI-related work is happening across UCSB in many forms (academic programs, research centers, administrative tooling, teaching innovation, student services, and public-facing communication). There are multiple examples within Academic Affairs alone: Individual units are pursuing meaningful work: CITRAL is developing faculty resources, the Computer Science department has implemented an AI degree, the library is offering workshops and courses in AI literacy including vibe coding, and an AI Speaker Series, IT is piloting Gemini and an AI sandbox that includes multiple models (Gemini, Claude, Amazon, Qwen, Llama), and multiple research groups are advancing AI-related scholarship. The AI Community of Practice has created an informal gathering space that has generated real cross-campus conversation. These efforts could benefit greatly from cross-coordination or even collaboration, but are currently largely disjointed. Units are not aware of what other units are doing, are not cross-promoting related initiatives, and are not building on each other’s work. The result is duplication, missed opportunities for collaboration, and a public profile that does not reflect the breadth or seriousness of what is actually happening at UCSB. This is a structural gap. No group has been assigned responsibility for seeing the whole picture, connecting the pieces, or representing UCSB’s AI activity in a coherent way. The answer to “Who is leading for AI on campus?” is currently everyone and no one.

Details of Recommendation
Integrate AI into the campus strategic planning process. The committee understands that the Chancellor will be entering a period of institutional strategic planning. The committee’s strong feeling is that AI must be a meaningful part of whatever campus strategy emerges from that process. AI touches virtually every aspect of the university’s mission: teaching, research, administration, public engagement, and workforce development. It should not be siloed as a technology initiative or treated as a standalone section of the plan, but rather integrated throughout as a core aspect of how the university pursues its broader goals. The committee recognizes that it cannot prescribe exactly how AI should be incorporated into a strategic plan whose process has not yet been defined. What it can say is that any strategic planning effort that does not seriously engage with AI will produce a plan that is incomplete on arrival.

Assign AI leadership and coordination involving the Senate and the administration. The committee’s primary recommendation is that the campus identify and empower a person or group with explicit responsibility for coordinating AI-related activity across academic, administrative, and research domains. This role should have sufficient visibility and authority to convene stakeholders, identify gaps and overlaps, and represent the campus AI landscape both internally and externally. The committee does not prescribe the exact form this should take. It could be an individual appointment (an AI coordinator or similar role with cross-campus scope), a small coordinating body, or an expansion of an existing office’s mandate. What matters is that someone is clearly responsible for the big picture and has the standing to bring people together across organizational boundaries. The position should address both academic and administrative interests in aligning AI with the campus’s mission/vision, governance/policy, ethics/principles, research, and education.

Build a coherent AI-related public profile for the campus. The committee discussed the need for a more coherent AI public profile, specifically a communication strategy that represents UCSB's strengths in AI research, teaching, public outreach, and partnerships. Such a profile will help to attract students, faculty, and partners, and position the campus strategically and appropriately relative to emerging technology.
However, the committee also concluded that such a communication strategy cannot precede the development of an actual campus AI strategy. A webpage or a set of messaging themes, without someone setting the agenda and maintaining the big picture, will be out of date almost immediately. Thus, the public profile should emerge organically from the long-term strategic planning and coordination work: once someone can see what is happening across campus, articulating it becomes straightforward.

The committee discussed the value of being able to articulate UCSB’s AI posture in a few clear, action-oriented terms (verbs rather than nouns) that capture what the campus is trying to do with AI rather than what structures it has built. Whether those terms are something like “explore” or “accelerate” or something else, the point is that a simple, shared vocabulary makes coordination and communication dramatically easier. Developing that vocabulary should be an early priority for whoever takes on the coordination role.

Key Considerations
This is both structural and relational. The coordination function needs formal standing (structural), but its effectiveness will depend on relationships, trust, and the ability to convene people who may not otherwise talk to each other (relational). A dotted-line organizational chart alone will not solve the problem.

The academic and administrative dimensions are different but connected. AI in teaching and research raises different questions than AI in administrative operations or campus IT. Any coordination structure needs to hold both dimensions without collapsing them into one. The academic side involves both research strategy and new collaborations, but also includes faculty governance, academic freedom, and disciplinary variation. The administrative side involves data governance and operational efficiency. All of the above need attention, and they need to be connected.

Existing work should be built upon, not duplicated. The AI Community of Practice, CITRAL’s faculty development work, IT’s piloting efforts, and this committee’s own recommendations represent a significant body of work. The coordination role should amplify and connect these efforts, not create parallel structures that compete with them. The committee’s own role is time-limited. This advisory committee was charged with developing recommendations. The coordination and leadership function the committee is recommending is ongoing and operational and extends beyond what an advisory committee can or should do. Establishing that function is how the committee’s work gets carried forward.

Complicating Factors
There is no obvious institutional home for this function. AI does not fit neatly within IT, within the Academic Senate, or within any single college or academic unit. The coordination role will need to bridge all of these, which makes placement and reporting lines genuinely difficult. The committee does not view this difficulty as a reason to defer action.

Resources are constrained. Creating a new position or office requires funding. However, the cost of continued fragmentation (in duplicated effort, missed collaboration, and an incoherent public presence) is also real, even if harder to measure.

The landscape is changing rapidly. Any coordination structure must be designed for adaptability rather than permanence. The specific questions that matter in 2026 (model access, detection tools, course policies) will not be the questions that matter in 2029. The structure should be designed to evolve with the technology and the institution’s understanding of it.