School/AI Foundations/Core Concepts
3/4
Wave 110 minbeginner

AI Ethics & Limitations

What AI gets wrong, where it's biased, and how to use it responsibly.

AI Ethics & Limitations

AI is powerful, but it comes with real risks. Being an effective AI user means knowing the boundaries.

Known Limitations

1. Hallucinations

AI confidently states things that are completely false. It doesn't "know" when it's wrong because it's generating statistically likely text, not looking up facts.

Real examples of hallucinations:

  • Citing academic papers that don't exist
  • Making up court case precedents (this got a lawyer sanctioned)
  • Providing incorrect code that looks correct
  • Fabricating statistics

Defense: Always verify facts, especially for anything you'll publish or act on.

2. Training Data Cutoff

Most models have a knowledge cutoff date. They don't know about events after their training data was collected.

Defense: For current events or recent information, use AI search tools (Perplexity) or verify with up-to-date sources.

3. Bias

AI reflects biases in its training data. This can manifest as:

  • Gender stereotypes (assuming doctors are male, nurses are female)
  • Cultural bias (defaulting to Western/American perspectives)
  • Representation gaps (performing worse for underrepresented groups)

Defense: Be aware, review outputs critically, and explicitly ask for diverse perspectives.

Ethical Guidelines

DO:

  • Disclose AI use when appropriate (especially in academic or professional settings)
  • Verify important claims before acting on them
  • Use AI to augment your work, not replace your judgment
  • Protect sensitive data — don't paste passwords, SSNs, or private medical info

DON'T:

  • Submit AI-generated work as entirely your own in contexts where that's dishonest
  • Use AI to generate misinformation or deceptive content
  • Blindly trust AI for medical, legal, or financial advice
  • Share others' private information with AI tools

The Copyright Question

AI-generated content occupies a legal gray area. Currently:

  • You generally can't copyright purely AI-generated content
  • Using AI as a tool in your creative process is fine
  • The law is evolving rapidly — stay informed

Exercises

0/3
Quiz+5 XP

A lawyer was sanctioned for using AI because:

Prompt Challenge+10 XP

Ask AI: "Who won the Nobel Prize in Literature last year?" and then verify the answer with a web search. Was the AI correct?

Hint: This tests the training data cutoff. The AI may give an outdated or fabricated answer.

Quiz+5 XP

Which is the BEST practice when using AI for work?