suggested revisions to the AI for Code Development page

[codebeaker]

URL of existing content (a rendered, sciwiki.fredhutch.org link)

https://sciwiki.fredhutch.org/datascience/ai_coding/

Specific information requests or concerns, or specific region on page of content to refresh.

Thoughts and questions I had reading this draft:

1. Where it says to "Choose enterprise-supported or locally hosted AI models rather than public, cloud-based tools that may store or log sensitive code or data." -- I wanted an action to take. Like examples of what tools are ok to use, or where to find a list of acceptable tools that are maintained, or even just a reminder that I can email dataprotection@fredhutch to ask about this (even though it literally says that right above).

2. There's one use case that's not listed as do or not do that maybe should be, which is putting your code into an LLM for troubleshooting. It is related to the suggested use 'suggest performance improvements to existing code' but it might still be worth calling out specifically.

3. We tell them not to assume AI-generated code is secure but then give as a suggested use case that they have AI generate secure code or security checklists. Perhaps we add a line to those bullet points reminding them to review the code? e.g. "Prompt AI to generate security-hardened code (e.g., “Write a secure API endpoint using OAuth2 authentication”). Remember to review all AI-generated code to ensure you understand what it is doing before you execute it"?

@ GitHub username of the listed primary reviewer of the page if known

@vortexing