Rice University leaders from across campus emphasized the importance of balancing artificial intelligence innovation with privacy, security, accessibility and institutional accountability during an AI Week panel discussion Oct. 22. It was one of more than 30 panels conducted during Rice’s inaugural AI Week that centered on the use of AI technology in academia and bolstered the university’s strategic plan Momentous, which calls for responsible AI.
The discussion, “Rules of the Road: Building and Implementing AI Responsibly Across Campus,” brought together representatives from compliance, information security, student records, disability services, equal opportunity and legal affairs to examine how AI is reshaping higher education and what responsible adoption looks like in practice.
Moderated by Omar Syed, vice president and general counsel, the panel featured Chetna Koshy, associate vice president and chief compliance officer; Kim Woodruff, director and deputy chief of the Information Security Office; Justin Schilke, university registrar; Sandra Harrison, director of the Student Disability Resource Center; and Richard Baker, executive director for Equal Opportunity Services and Title IX and ADA coordinator.
Throughout the discussion, panelists returned to a common theme: Responsible AI should enable innovation, not impede it, while ensuring that new technologies align with Rice’s academic mission, institutional values and obligations to its community.
“Responsible AI” means more than complying with existing laws and policies, Koshy said. It also requires transparency, consent, preservation of intellectual work and human accountability.
Rather than relying exclusively on a fixed AI policy that could become outdated as technology advances, Rice’s AI advisory committee is developing guiding values intended to help the university respond to emerging technologies and risks. Koshy encouraged faculty, staff and students to review the proposed values and provide feedback.
The goal is “to see how we can support our faculty, our staff and our students to really boldly innovate without concern of risk to our academic mission or our institutional trust that we built for over a century,” Koshy said.
Protecting data in an increasingly AI-enabled environment
Information security and data protection emerged as central concerns, particularly as employees and researchers gain access to a growing range of AI tools.
Woodruff cautioned that consumer AI subscriptions may not provide the contractual protections necessary for university information. Users should understand how a tool handles submitted data, whether information may be used to train models and what protections apply to both inputs and outputs.
Rice is working to expand access to enterprise AI tools with appropriate contractual safeguards, she said. In the meantime, university technology-use and data-protection policies continue to apply to AI, and users must consider the sensitivity of the information they share.
Woodruff also highlighted emerging cybersecurity threats, including increasingly sophisticated phishing attempts, AI-assisted discovery of software vulnerabilities and risks associated with AI agents interacting with applications and data. AI-generated information can also be inaccurate, even when a system appears confident.
Her message to the campus community was direct: Understand the tool, understand the data and verify the results rather than trusting AI automatically.
Schilke addressed the implications for student privacy under the Family Educational Rights and Privacy Act. As AI becomes more involved in education, student records may include not only traditional documents such as transcripts and graded assignments but also AI-generated feedback, learning analytics and automated assessments.
He emphasized that the use of AI does not eliminate students’ rights to access and seek amendment of applicable education records. Institutions must therefore consider whether they can explain how AI-assisted decisions are made and provide meaningful transparency about the information used.
Schilke advocated a “privacy by design” approach that includes data minimization, data mapping, access controls, ongoing monitoring, staff training and incident response procedures.
He also warned that removing names and other direct identifiers from information does not necessarily make it anonymous. AI systems may combine data from different sources in ways that make individuals identifiable again.
Schilke’s takeaway for faculty and staff: Protect student information before it enters an AI system, and ensure that the tool is appropriate for the data and intended use.
Fairness and accessibility must be built in
Panelists also examined how AI can reproduce existing biases and create barriers for people whose experiences are not adequately represented in the data used to develop AI systems.
Baker discussed the risks of AI-assisted hiring, resume screening and employee evaluations. Systems trained on historical employment data, for example, may reproduce past patterns of discrimination. Automated assessments may also disadvantage people because of disability-related communication differences, age, pregnancy-related leave or other factors unrelated to their qualifications.
An organization cannot simply transfer responsibility to a technology vendor or an automated system, Baker said. Human decision-makers must independently evaluate the reasons behind consequential decisions and remain accountable for their outcomes.
He also urged institutions to consider the broader effects of workplace automation and preserve opportunities for meaningful human interaction.
Harrison brought an accessibility perspective to the discussion, emphasizing the importance of including people with disabilities in AI development and testing. Models designed around the experiences of a majority population may not adequately serve people with nonstandard speech, cognitive differences or diverse physical behaviors.
At the same time, AI can offer substantial benefits for students with disabilities.
AI-supported note-taking and transcription can help students who have difficulty taking notes while listening to a lecture. Vision-assistance tools can help users read signs and navigate unfamiliar environments. Planning and organizational tools can support students who experience executive-function challenges.
These applications demonstrate how AI can expand access to education and independence when designed around the needs of diverse users.
Harrison encouraged the university community to consider who might be excluded from a technology’s benefits and to involve accessibility and equal opportunity experts early in the development process.
Koshy reinforced that recommendation, noting that early consultation allows potential risks to be identified and mitigated before an AI system is deployed.
Innovation with appropriate safeguards
Despite the risks, panelists highlighted opportunities for AI to improve teaching, research and university operations.
Schilke described how AI can support student success through personalized learning pathways, earlier identification of students who may benefit from academic advising and improved administrative processes. He cited exam scheduling as one example: AI can help distribute final exams more effectively, so students face fewer clusters of exams.
Harrison’s examples illustrated how AI can support accessibility, while other panelists pointed to opportunities to improve efficiency and help faculty, staff and researchers focus on work that benefits from human expertise.
The challenge is to ensure that these gains do not come at the expense of privacy, fairness or trust.
Panelists also discussed the evolving regulatory landscape. Although the United States does not have a single comprehensive federal AI law governing all uses of the technology, existing laws and contractual obligations may still apply. Privacy, nondiscrimination, accessibility and information-security requirements remain relevant, as do restrictions attached to certain sponsored research projects.
For research funded by the Department of Defense or other federal agencies, Syed advised researchers to consult Rice’s Office of Sponsored Projects and seek legal guidance when necessary to determine which AI tools are permitted under the applicable agreements.
Building a campus culture of responsible AI
In response to an audience question about whether Rice prohibits nonenterprise AI chatbots, Woodruff and Koshy clarified that the university has not adopted a blanket campuswide ban. However, specific research contracts may restrict the use of particular tools, and existing university policies govern the appropriate handling of information.
Woodruff also announced plans for an interactive AI advisory tool intended to help members of the Rice community identify appropriate AI tools based on the type of information and research involved. The tool was described as an upcoming resource at the time of the panel.
The discussion concluded with a question about AI’s environmental impact, including the resources and energy required to operate these technologies. Panelists identified sustainability as an important topic for further consideration.
Ultimately, the panel underscored that responsible AI adoption requires ongoing collaboration among the university’s stakeholders.
As AI capabilities continue to evolve, Rice’s approach will depend on a combination of adaptable guiding principles, practical safeguards and informed human judgment.
The goal is not simply to use AI more widely but to use it in ways that strengthen the university’s academic mission while protecting the people, information and institutional trust that make innovation possible.