Guidance on the Use of Generative AI and Large Language Model Tools
To: The Caltech Community
From: David A. Tirrell, Provost
Charles E. Lane, Vice President and Chief Operating Officer
We write to provide initial guidance to encourage the responsible use of generative artificial intelligence (GenAI) and large language model (LLM) tools and technologies, such as OpenAI's ChatGPT and Dall-E and Google's Bard in research, education, and administrative work at Caltech.
As a research and education institute, committed to advancing the frontiers of science and engineering and expanding knowledge, we support a responsible, measured experimentation with and use of new technologies. While doing this, however, Caltech requires that you follow all existing applicable regulations and Institute policies. These include, but are not limited to, ensuring protection of confidential, personal, or business information and intellectual property, and adherence to the honor code, course requirements, research integrity, and publication ethics.
GenAI and LLM technologies have evolved rapidly in their use and application this past year and are expected to continue to evolve in ways society cannot predict. Likewise, our guidance for the appropriate use of these tools is written for the present moment and will likely evolve alongside the technology. In the interim, however, as you use GenAI and LLM technologies in your work at Caltech, we ask that you apply these four guiding principles to your practice: disclosure, data and information protection, content responsibility, and Caltech's honor code.
- Disclosure: When using GenAI, always disclose promptly, or reference the use of GenAI tools and application plug-ins, as applicable. This transparent disclosure ensures that others are aware when GenAI was used to generate content and reduces misunderstandings regarding the source of information, potentially limiting claims of academic dishonesty or plagiarism. When using GenAI to write or publish, please make sure to follow the guidance provided by the course instructor or journal or manuscript publisher/editor. For example, some may require that the GenAI be included as an author, others may simply require acknowledgement.
- Data and Information Protection: Federal, state, and local laws as well as Caltech policies may limit data that can be disclosed. Unless you are using a GenAI application that ensures separation of your entry from other entries and confidentiality (usually a paid service), uploading content into GenAI (Open GenAI) is a public disclosure. It is safest to assume data or queries uploaded into Open GenAI tools will become public information, unless otherwise indicated.
In order to protect Caltech data and information, do not enter, contribute, or otherwise input sensitive, confidential, or restricted information into Open GenAI tools. This includes, but is not limited to, data covered by regulations such as FERPA and HIPAA, any intellectual property or unpublished research data, export-controlled data, and other sensitive HR, business, or administrative data. Caltech is considering a subscription to a restricted GenAI and will keep you apprised of its progress in securing such a service. In the meantime, Caltech has and continues to reserve the right to disable or limit access to AI companion tools in enterprise business software and applications, such as Zoom and Microsoft Office suites.
- Content Responsibility: Remember that GenAI systems are fallible. Responses can be inaccurate, misleading, and even entirely fabricated. Therefore, you should always review and assess all output generated by GenAI tools for accuracy before relying on them or distributing them publicly.
- Honor Code: Caltech's honor code underscores the importance of ethical conduct and fairness and extends to the use of GenAI tools and is stated as follows: "No member of the Caltech community shall take unfair advantage of any other member of the Caltech community." These guidelines have been posted to the Information Management and Support Services (IMSS) website. Please note that we offer these guidelines in addition to the teaching resources that have already been provided by the Center for Teaching, Learning, and Outreach.
The Institute's guidance promotes responsible and ethical use of GenAI and LLM tools at Caltech, and fosters a community that values transparency, integrity, privacy, accuracy, and fairness. Caltech may update its guidance as the technology and regulatory and commercial landscapes evolve. Thank you for adhering to these guidelines. If you have any questions about the use of this dynamic technology, please email [email protected].