Why you need this
AI can amplify bias, leak data, and cause real harm if deployed carelessly. This checklist helps you catch ethical issues before they become crises.
Use it before launching AI features, during audits, or as part of your governance process.
Perfect for:
- Product teams launching AI-powered features
- CTOs and engineering leaders setting AI policies
- Compliance officers auditing AI systems
- Startups that want to build responsibly from day one
What's inside
10 Essential Ethics Checkpoints
Each checkpoint includes:
- The question — Yes/no format for easy auditing
- Why it matters — Consequences of ignoring this issue
- Action items — Concrete steps to address gaps
1. Privacy & Data Protection
Question: Have you identified and protected all sensitive data?
- What counts as sensitive (PII, health data, financial records)
- How to anonymize or exclude risky data
- Privacy-by-design principles
2. Bias & Fairness
Question: Have you tested your AI for bias across demographics?
- Where bias shows up (hiring, lending, content moderation)
- How to audit outputs for discriminatory patterns
- Strategies to reduce bias (diverse training data, human review)
3. Transparency
Question: Do users know when they're interacting with AI?
- Disclosure requirements (when to say "this is AI-generated")
- Explainability: can you explain how decisions are made?
- Avoiding "black box" systems for high-stakes decisions
4. Accountability
Question: Is there a human responsible for AI outputs?
- Who reviews high-stakes decisions?
- What happens when AI makes a mistake?
- Clear escalation paths for errors or harm
5. Accuracy & Reliability
Question: Have you measured and disclosed error rates?
- Hallucination rates and how to communicate them
- When to require human verification
- Setting user expectations for AI limitations
6. Consent
Question: Did users consent to AI use of their data?
- Opt-in vs opt-out considerations
- Clear terms of service and privacy policies
- Compliance with GDPR, CCPA, and other regulations
7. Safety & Harm Prevention
Question: Have you implemented guardrails to prevent misuse?
- Content filters for harmful outputs
- Rate limiting and abuse detection
- Emergency shutdown procedures
8. Environmental Impact
Question: Have you considered the carbon footprint of your AI?
- Training vs inference emissions
- Model size optimization
- Choosing efficient models and infrastructure
9. Accessibility
Question: Is your AI accessible to people with disabilities?
- Screen reader compatibility
- Alternative input methods
- Bias against non-standard language or dialects
10. Continuous Monitoring
Question: Do you have a process for ongoing audits?
- Regular bias testing
- User feedback loops
- Incident response plans for ethical violations
How to use it
For Product Launches:
- Run through the checklist before going live
- Document gaps and create remediation plans
- Share results with leadership and legal teams
For Audits:
- Use as a quarterly review framework
- Compare results over time to track improvement
- Identify patterns (e.g., "we consistently miss on accessibility")
For Training:
- Walk through the checklist with your team
- Discuss real-world examples of each issue
- Create team-specific action items
Real-world example: Hiring AI
Scenario: You're building an AI to screen job applications.
Checklist results:
- ✅ Privacy: Candidate data is encrypted and anonymized
- ❌ Bias: Not tested across demographics → Action: Audit for gender/race bias
- ✅ Transparency: Candidates are told AI is used
- ❌ Accountability: No human reviews AI rejections → Action: Require HR review
- ✅ Accuracy: Error rates disclosed to hiring managers
- ✅ Consent: Privacy policy covers AI use
- ✅ Safety: No harmful outputs possible (it's just scoring)
- ✅ Environmental: Using efficient model (GPT-3.5)
- ❌ Accessibility: Doesn't handle non-English names well → Action: Test with diverse names
- ✅ Monitoring: Monthly bias audits scheduled
Result: 7/10 passed. Clear action items to fix gaps before launch.
Want to go deeper?
This checklist is your starting point. For detailed guidance on each topic:
- AI Safety Basics — Privacy, verification, and policy guidelines
- Glossary: Bias — Understanding AI bias and fairness
- Glossary: Hallucination — AI reliability issues
License & Attribution
This resource is licensed under Creative Commons Attribution 4.0 (CC-BY). You're free to:
- Share with your team or organization
- Print for audits or compliance reviews
- Adapt for your industry's specific requirements
Just include this attribution:
"AI Ethics Checklist" by Field Guide to AI (fieldguidetoai.com) is licensed under CC BY 4.0
Download now
Click below for instant access. No signup required.