Generative AI Guidelines

Recent advances in generative artificial intelligence (AI) tools have broadened our capabilities and spurred endless conversation about the benefits and risks. Is AI taking our jobs? Is AI biased? Can we trust the results?

Generative AI is a convenient tool for brainstorming, research, and completing a task quickly — however, it requires human oversight. AI is quick, but it lacks humanity. Care should always be taken to avoid letting the AI do the thinking, the feeling, or the persuading.

Protecting the University’s reputation, the integrity of our data, and the privacy of our community should drive every decision to use artificial intelligence in our work. To that end, University Communications has crafted guidelines for proper use of AI at UConn.

Data Security Consideration

Any confidential, protected, proprietary, or embargoed information, images, logos, and data must never be entered into an AI platform. Sharing content with large language models, such as ChatGPT, exposes that content to others outside of the University and could be stored externally and used to create unlicensed content in the future. Carefully consider potential consequences before submitting any proprietary creative content.

Please note that Copilot for Microsoft 365 integrates with core Microsoft 365 applications, but as long as users are logged in, the responses are not used to train underlying AI models. Read more via ITS.

Review, Verify, Modify, and Attribute

While UConn may sometimes use AI, we adhere to these guidelines:

Review. All AI-derived content will be reviewed by a human.
Verify. While fast, AI has shown a tendency to make things up. A human must fact-check AI results for accuracy and bias.
Modify. Any written, image, or video content produced even in part by AI will be modified by a human author before being used in any public-facing context. Ensure UConn’s brand tone and voice are reflected in final content.
Attribute. Material quoted or paraphrased from any source must be cited with attribution to avoid plagiarizing. This includes AI.

When To Use AI

Before enlisting AI’s help, it’s useful to ask: What is the goal?

In written communications, if you’re looking to persuade, evoke an emotion, or connect with your reader, then write a first draft yourself. That first draft will have more personality in it when it matters most. If, however, you’re seeking a collection of information that you will use as research or as an outline, then asking AI first could be useful.

When it comes to visual content, the rules are more specific. To spark an idea or to literally see something differently, generative AI can be useful — as long as you’re considering UConn’s privacy and data security rules above.

If you’re still unsure, please reach out to Craig Burdick in University Communications.

Acceptable Uses of AI

Research
- Preparation. AI can be used to learn background on story topic, to help prepare interview questions for reporting, to find sources, to help understand complex topics, etc.
- Brainstorming. Use AI to generate new and creative ideas. These can be for any kind of content, like stories, press releases, social media posts, and website copy. (Do not, however, expect AI-generated content to be free of errors. Always verify.)
- Contact information. AI can be used to quickly find and organize contact information for people, such as story sources.
Writing and editing
- Planning and organization. Organize content into outlines, editorial plans, or story summaries for different platforms.
- Editing and revisions. AI can assist with grammar, sentence structure, tone, and style revisions, and function as a thesaurus.
- Search Engine Optimization (SEO). Check for keyword optimization and enhance readability to improve SEO (e.g., headlines, tags).
- Accessibility. Support implementation of accessibility like generating alternative text for images.
Web content and management
- Content analysis. AI can optimize human-generated web content for SEO, accessibility, and performance.
- Content structuring. Generate headings, subheading ideas, and structural outlines tailored for specific audiences.
Social media management
- Post ideas. Brainstorm ideas for social media posts and campaigns.
- Audience analysis. Utilize AI for social listening, trend analysis, and generating insights about audience behavior.
- Content drafting. Draft social media content with audience-specific optimization.
- Tagging. Find suggestions for useful or popular hashtags.
Creative visual content
- Photo editing. Use AI tools for editing tasks like color correction, noise reduction, and composition improvements without altering the original intent.
- Visual inspiration and concepting. Generate mood boards, concept art, or abstract illustrations to guide visual creation.
- Image optimization. Extend or enhance images for new formats, ensuring respect for intellectual property laws.
Data and analytics
- Data gathering. Assist in competitive analysis, market research, and campaign performance reviews.
- Engagement. Analyze metrics to gain insights — for example, on social media engagement and content performance.
Routine tasks and productivity enhancements
- Content Summaries. Summarize meeting notes, interviews, transcripts, etc.
- Draft Routine Communications. Assist with drafting emails, presentations, or reports.
- Content Organization. Help manage editorial calendars and content strategies.

Note: All written work that will be published for the public should be scrutinized for errors, tone, and readability. Humans should still streamline content and perform the last review, editing and tailoring steps before publication.

Unacceptable Uses of AI

Research
- Do not share sensitive, confidential, or embargoed information with an AI tool, which can include student, staff, or patient information (e.g., FERPA, HIPAA) or information the University has promised to keep private.
Writing and editing
- UConn does not support publishing content that is completely AI-generated.
Web content and management
- UConn does not support publishing content that is completely AI-generated.
- Do not copy and paste the results of a ChatGPT prompt directly onto a UConn website without first verifying its accuracy, checking for typos or other mistakes, and ensuring it adheres to UConn’s brand standards.
Social media management
- Do not copy and paste the results of a ChatGPT prompt directly onto a social media platform without first verifying its accuracy, checking for typos or other mistakes, and ensuring it adheres to UConn’s brand standards.
- If using generative AI to create an image from scratch, do not upload Jonathan the Husky in any form (e.g., canine, mascot, logo); share logos, wordmarks, or trademarked content, including UConn trademarks; or intend to imply AI-generated artwork is reality.
- Before sharing any AI-generated images, make sure they do not contain another school’s or company’s trademarked logos or identifying marks. UConn is not the only school with a husky for a mascot.
Creative audio and visual content
- When employing AI’s help, do not upload Jonathan the Husky in any form (e.g., canine, mascot, logo); share logos, wordmarks, or trademarked content, including UConn trademarks; intend to imply AI-generated artwork is reality; or use AI tools to voice-clone or deep-fake audio or video.
Data and analytics
- Do not share sensitive, confidential, or embargoed information with an AI tool, which can include student, staff, or patient information (e.g., FERPA, HIPAA) or information the University has promised to keep private.
Routine tasks and productivity enhancements
- Do not share sensitive, confidential, or embargoed information with an AI tool, which can include student, staff, or patient information (e.g., FERPA, HIPAA) or information the University has promised to keep private.

Contact

Adapting to ever-changing technology like generative AI takes time and dedication. Contact University Communications with questions about how and when to use AI in your work.