AI ethics asks how systems should be built and used: who benefits, who bears risk, and what guardrails prevent harm. Technical teams share responsibility with legal, policy, and domain experts—not only after launch.
Why ethics is not optional
- Automated decisions scale mistakes and bias
- Opacity erodes trust when errors occur
- Regulators and customers increasingly expect documentation
- Reputational and legal costs exceed model savings when things go wrong
Stakeholders
Users, affected non-users (e.g., people scored without consent), workers displaced or surveilled, society at large. Impact assessment should include voices beyond the paying customer.
Ethics checklist seed
questions = [
"Who could be harmed if this fails?",
"Do we have consent and lawful basis for data?",
"Can a human override the model?",
]
for q in questions:
print("-", q)Practice: Reflect on ethics scenarios in writing—no code required. Optional snippets illustrate policy checks only.
Important interview questions and answers
- Q: Ethics vs compliance?
A: Compliance meets law minimum; ethics asks what ought to be done when law is silent or lagging. - Q: When involve legal?
A: Before collecting sensitive data, cross-border transfer, or automated decisions with legal effect.
Self-check
- Name two reasons ethics matters for AI products.
- Who are stakeholders beyond paying users?
Tip: Run the three harm/consent/override questions at kickoff—not post-launch.
Interview prep
- Ethics vs compliance?
- Compliance meets legal minimum; ethics asks what should be done beyond lagging law.
- Stakeholders beyond users?
- Affected non-users, workers, and broader society.