ITS Guidelines for Generative AI Use

Recommended AI Tools

• For general use, ITS recommends use of Google Gemini for all employee and student needs.

  • Gemini is the only AI tool available to the full NCSSM community that:

    • Does not use prompts or data inputs to train models

    • Does not allow for human review by Google

    • Enables ITS’ full support

  • When in doubt use Gemini and other Google AI apps, such as Gemini Notebook, in conjunction with your NCSSM account

  • For any other AI tools or applications that utilize AI features contact ITS for review and approval

• For meeting transcription, ITS recommends use of Zoom AI Companion for all employee needs.

  • Many AI-based transcription tools exist

  • The only supported AI-based transcription tool supported for employees is Zoom AI Companion

    • Every faculty and staff member with a Zoom license from ITS is already licensed for, and can enable, Zoom AI Companion

    • Using this tool enables the best possible privacy, security, and support

  • If you want to integrate any other AI tools, you are required to contact ITS for review and approval

Data privacy and security

• Protect sensitive data by not entering internal data (classified as Level 1 and above) into publicly-available generative AI tools. Information shared with generative AI tools is typically not private and could expose sensitive information to unauthorized parties.

Exercise caution about content

• AI-generated content can be inaccurate, misleading, or entirely fabricated (sometimes called “hallucinations”), or may contain copyrighted material. Review your AI-generated content before sharing it.

• If you come across any AI-related concerns or issues, whether related to privacy, bias, or ethical considerations, please report them to the appropriate channels such as ITS, Institutional Effectiveness, or Academic Programs.

• Text and images generated by AI may be used for phishing and other social engineering attacks, so please be extra vigilant about verifying those asking for information and follow recommended best practices to report anything that appears suspicious.

Mitigating bias

• Be mindful of potential biases in AI systems and work to mitigate them, especially when developing or using AI applications. Bias can be a significant issue with generative AI, as it may inadvertently create content that reflects societal biases. To reduce this bias, it is valuable to know the source of a model’s training data, how humans participated in the training process, the steps that were taken to measure and reduce bias, and how the model is updated. To help review this information for many generative AI models, Stanford researchers have created a comprehensive transparency index.

AI detection tools (don’t use them)

• AI detection tools and software are known to have significant deficits and are not recommended for employee use.

  • Many false positives and false negatives

  • May be biased against certain students’ writing styles

  • Cannot be used as evidence for conduct and discipline

• Instead of relying on these types of tools, a more holistic approach is needed.

Academic use

• A committee representing the faculty on both campuses has shared the following guidelines: https://docs.google.com/document/d/1PH1if14c5qbleAEj2ySn595ZmmJpjZTHR29xDfu2fSQ/edit?usp=sharing

• In summary, this outlines guidance on AI tool usage for students, emphasizing:

  • AI is generally not permitted for assignments unless specifically allowed by an instructor.

  • Instructors will indicate permitted uses (categorized into brainstorming, editing, feedback, instruction & tutoring, media generation, summarization, translation, and writing) within syllabi or on individual assessments, providing clarification for nuanced applications.

  • When AI tools are allowed, students are fully responsible for the generated material, must ensure accuracy, and are required to cite all AI-generated content according to MLA or APA guidelines, or as directed by the instructor.

  • If unsure about AI tool usage, students must seek clarification from their instructor.

Creating apps and tools ("vibe-coding")

A number of NCSSM community members use AI to build custom automations or applications to improve workflows. To ensure these internal tools are secure and sustainable, ITS requires a structured partnership. If you plan to develop a tool using AI, please adhere to the following:

• Submit a request: Before you start building, you must submit a Software Development Partnership Request via the ITS Help Desk. Approvals are based on the classification level of the data your tool will access.

• Verify accuracy: Fully unsupervised AI tools (a "black box" implementation) are prohibited. You must thoroughly test your tool to ensure that your use of AI does not "hallucinate" or invent data.

• Protect institutional data: AI agents must be strictly "read-only" and must interact with data via ITS-approved APIs. Any "write" access to institutional data requires explicit ITS authorization and strict operational limitations.

• Ensure long-term support: Your divisional or departmental leadership must approve the tool, identify a secondary owner for succession planning, and save all code and development documentation (such as system prompts, model versions, and changes) in a centralized ITS repository rather than a personal drive.

• References and additional details on requirements:

  • https://drive.google.com/file/d/1JPC2dXgu8sPNlC5JID1L6Xtfr9DjogH7/view

  • https://drive.google.com/file/d/1DtQGo9s4ZmcOz4L4iK3m0_yTCZvNMd42/view

We encourage all employees to approach generative AI tools with a thoughtful and responsible mindset. These tools offer substantial benefits to our campus community, but they also require careful consideration for their use.

Should you have any questions or concerns about software or services utilizing generative AI, either on an institutional device, the school network, or in any aspect of your work at NCSSM, please feel free to contact ITS at ithelp.ncssm.edu or via email at ithelp@ncssm.edu.

Should you consider using an generative AI-based tool in conjunction with institutional data, please submit a data request here or by email at datarequest@ncssm.edu.