GenAI Do's and Don'ts
What to do (and not to do!) when accessing and utilizing AI tools and resources.
Prompt Engineering
Create new chats for new topics to keep conversations organized.
Use follow-up prompts in the same chat thread for continuity.
Be specific: provide context, structure, tone, and sample writing styles.
Provide information about your audience so the response can be tailored to resonate with them.
Use the "regenerate" feature to have the tool try again if it misses the mark.
Don't treat it like a search engine — responses are generated based on patterns and probabilities derived from a trained model, not direct links to sources.
Don't assume all output is factual — always verify information.
Don't use it for math — it has difficulty performing mathematical calculations.
Don't include sensitive or personally identifying information.
Don't expect perfection — treat responses as drafts and refine them as needed.
General Use
Choose Wisely: Consult the UC-licensed GenAI tools to determine which tool to use with special consideration for the approved use by privacy level.
Stay Informed: Complete recommended UC GenAI training programs and regularly review updated policies and ethical guidelines.
Maintain Transparency: Clearly disclose when GenAI has been used in content creation.
Apply Human Oversight: Treat GenAI as an assistant, not a replacement. Always review, edit, fact-check, and approve AI-generated content before use.
Use AI for Low-Risk Tasks: Employ GenAI for brainstorming, summarizing documents, improving writing clarity, and generating non-sensitive content.
Consider Environmental Impact: Use GenAI efficiently, minimizing high-energy or resource-intensive requests like excessive image generation. Use traditional search engines whenever possible.
Never Enter Sensitive Information: Never input confidential, sensitive, or UC Berkeley proprietary data into GenAI tools.
Don’t Assume AI is Always Accurate: GenAI can generate incorrect or misleading information; always verify outputs.
Don’t Use AI to Generate Misleading Content: GenAI should never be used to create deepfakes.
Don’t Rely on AI for Critical Decisions: Never use GenAI for high-stakes decisions like hiring, performance evaluations, or evaluating bid responses.
Don’t Use AI for Legal, Financial, or Compliance Matters: AI-generated content should not replace professional legal, financial, or policy-related guidance.
Don’t Create GenAI Models Without Approval: Any GenAI model development must be submitted through intake for governance review and must go through the university security review process.