When NOT to Use AI for Code

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-coding-when-not-to-use-ai-creators

1. A developer is building a healthcare application that must comply with HIPAA. They want to use a free, publicly-available AI coding tool to help with a feature that processes patient medical records. What should they do?

   1. Avoid AI entirely for healthcare projects
   2. Use the free tool since it works just as well for this task
   3. Use a business-tier tool with a data processing agreement that explicitly addresses HIPAA compliance
   4. Only use AI for the user interface, not the backend processing

2. Which scenario represents the strongest case for refusing to use AI coding tools?

   1. Developing an algorithm that has never been published before
   2. Creating a web scraping script for public data
   3. Writing cryptographic code that secures a financial institution's key management system
   4. Writing a simple function to calculate sales tax

3. A junior developer is learning a new programming language. They want to use an AI coding assistant to speed up their learning. What does the material recommend?

   1. Skip the AI and write code manually, because struggling through problems is essential to learning
   2. Use AI freely since it produces working code faster
   3. Use AI for half the tasks to balance speed and learning
   4. Only use AI for syntax help, not logic

4. A developer is working on a highly novel algorithm that could be a breakthrough in distributed systems. They consider using an AI coding assistant. What does the material advise?

   1. Avoid AI because it averages patterns and novel work is the opposite of average
   2. AI is perfect for this since it can generate complex code
   3. Use AI but verify every line manually
   4. AI should be used only for testing the algorithm

5. An engineer wants to paste proprietary competitor code into an AI coding tool to help understand and improve upon it. What is the primary risk?

   1. The AI will refuse to help with competitor code
   2. The code might be copyrighted and using AI violates fair use
   3. The code could appear in training data or logs and leak to competitors
   4. The AI might give wrong suggestions

6. A team is about to work on a legacy codebase that has zero automated tests. They want to use AI to help add new features. What should they do first?

   1. Use AI only for documentation, not code
   2. Add comprehensive tests before using AI, because you cannot verify AI output without tests
   3. Avoid the legacy code entirely and rewrite from scratch
   4. Start coding with AI since it produces working code

7. Which category of data should never be pasted into any AI coding tool, regardless of tier?

   1. Public library documentation
   2. Open source code from GitHub
   3. Customer names and email addresses
   4. API keys and database credentials

8. An organization implements several enterprise mitigations for AI coding tools. Which approach provides the strongest protection for highly sensitive proprietary algorithms?

   1. Using a self-hosted open-source model like Llama
   2. Using Copilot Business with default settings
   3. Using ChatGPT Enterprise with training disabled
   4. Using the free tier but not entering company names

9. A developer notices they have been relying on AI for almost all coding tasks for six months. They can now write basic code without AI but struggle to debug complex issues independently. What is this phenomenon called?

   1. Tool dependency syndrome
   2. Knowledge degradation
   3. Skill atrophy
   4. Debugging blindness

10. What is the current legal status of AI-generated code regarding copyright in the United States?

    1. AI-generated code is automatically copyrighted like human-written code
    2. AI-generated code becomes public domain after 90 days
    3. The copyright status is unsettled and may not be copyrightable if purely AI-generated
    4. AI-generated code has no copyright protection because it was created by a machine

11. A startup is building a commercial software product. They heavily used AI coding assistants throughout development. What should they document to protect themselves legally?

    1. The cost savings from using AI versus hiring more developers
    2. The names of all AI tools used during development
    3. Nothing special is needed since AI code is treated like human code
    4. Which parts are human-authored and preserve that provenance

12. Which of the following is listed as a 'soft no' category - situations where AI usually makes things worse?

    1. Writing documentation for existing code
    2. Performance-critical hot paths in a game engine
    3. Designing a company logo in an AI art tool
    4. Creating a new internal tool for the company

13. An enterprise wants to deploy AI coding tools company-wide. What should their policy explicitly include?

    1. A ban on using AI for any financial calculations
    2. Explicit allowed and denied categories along with a review cadence
    3. A requirement that all AI code must be reviewed by the CEO
    4. A list of allowed AI tools and password requirements

14. A developer is working on a very short task - a five-line function to validate an email address. What does the material recommend?

    1. Avoid AI for any task under 100 lines
    2. Use AI but only for short tasks to build trust
    3. Use AI since it can generate this instantly
    4. Type it yourself because prompt-and-review is slower than just typing it

15. A developer wants to paste customer addresses into an AI tool to help format them consistently. What category of data is this and why is it risky?

    1. PHI - protected health information is always illegal to share
    2. Internal data - only company secrets are protected
    3. Public data - no concern since addresses are public record
    4. PII - personal information raises privacy law and consent concerns