AI Ethics Cheat Sheet — Quick Rules for Right and Wrong AI Use
Status: 🟩 COMPLETE 🟦 LIVING Section: cheat-sheets Tags: ethics, cheat-sheet, responsible-ai, decisions, principles
How to read this
Practical ethical rules for AI use. Most are universal; some are Australian-specific. When you’re unsure whether AI use is ethical, run through these.
The five questions before any AI use
Before using AI for anything significant, ask:
1. Who benefits?
- You? Your client? Your employer? Everyone? Yourself only?
- Is the benefit fair?
2. Who could be harmed?
- Direct: specific people affected
- Indirect: trust, society, environment
- Potential vs likely
3. Am I being transparent?
- Would I be comfortable if everyone knew?
- Am I claiming AI work as fully my own?
- Are stakeholders aware?
4. Am I respecting consent and autonomy?
- People’s choices about their data
- People’s right to know
- People’s ability to opt out
5. Am I maintaining accountability?
- Who’s responsible if something goes wrong?
- Am I keeping records?
- Can I justify my choices?
If you can’t answer comfortably: reconsider.
Always disclose
You should always disclose AI use when:
- 🔴 Submitting work for assessment (school, university)
- 🔴 Publishing or broadcasting to public audience
- 🔴 Required by professional/regulatory standards
- 🔴 Required by your employer or client contract
- 🔴 The output would be assessed differently if known as AI
- 🔴 For commercial content where audiences expect human creation
- 🔴 In journalism contexts
- 🔴 For academic publishing
- 🔴 When asked directly
- 🔴 For AI-generated images of real people or situations
- 🔴 For AI-generated quotes or testimonials
- 🔴 For AI customer service (customers should know)
Disclosure formats:
- “This document was prepared with AI assistance for [purpose]”
- “AI-generated image” or “AI illustration”
- “Drafted with AI; reviewed and edited by [name]”
- “Generated using [specific tool]“
Never do (regardless of who’s watching)
- ❌ Submit AI work as fully your own when authenticity matters (academic, professional, creative)
- ❌ Fabricate quotes, sources, or citations
- ❌ Create deepfakes of real people without consent
- ❌ Clone voices without consent
- ❌ Generate fake reviews (illegal under Australian Consumer Law)
- ❌ Manipulate vulnerable people through AI
- ❌ Use AI to scam or defraud
- ❌ Generate child sexual abuse material
- ❌ Generate non-consensual sexual content of anyone
- ❌ Misrepresent humans as AI or AI as humans in ways that mislead
Things to verify before relying
- ✅ Every citation AI provides (frequently fabricated)
- ✅ Specific numbers and statistics (often hallucinated)
- ✅ Names and dates (frequently wrong)
- ✅ Legal claims (verify against authoritative sources)
- ✅ Medical information (verify with healthcare professional)
- ✅ Financial advice (verify with financial professional)
- ✅ Quotes attributed to people (often misattributed)
- ✅ Current information (AI has training cutoffs)
- ✅ Australian-specific information (often US-biased)
The professional responsibility principle
If you’re a professional (doctor, lawyer, accountant, engineer, etc.):
AI is a tool that may augment your work. AI is NOT a substitute for your professional accountability.
- You remain accountable for decisions
- Professional standards still apply
- Documentation requirements continue
- Indemnity considerations matter
- Client/patient interests come first
For Australian professionals:
- AHPRA for health practitioners
- CPA Australia / CAANZ for accountants
- Law Societies for lawyers
- Engineers Australia for engineers
- Professional bodies generally
All have AI guidance — read yours.
Privacy quick rules
Never put into AI tools:
- Passwords, credentials, 2FA codes
- Bank account numbers, full credit card numbers
- Tax File Numbers, Medicare numbers, passport numbers
- Other people’s identifying information without consent
- Confidential client/patient information (use enterprise tools with DPA)
- Sensitive personal information you wouldn’t share publicly
- Children’s identifying information
- Indigenous cultural information without protocols
Always consider:
- Where the AI provider stores data (Australia? US? China?)
- Whether they train on your data (most enterprise: no)
- Whether DPA covers Australian Privacy Act (APP 8)
- Records management obligations
The “harm to others” check
For content AI generates, ask:
- Could this be used to defraud someone?
- Could this damage someone’s reputation falsely?
- Could this be sexual content of a non-consenting person?
- Could this manipulate someone’s choices unfairly?
- Could this be used for harassment?
- Could this contribute to misinformation?
- Could this affect democratic processes?
If yes to any: don’t generate, or generate with significant safeguards.
For specific high-stakes contexts
Healthcare
- AI as decision support, never decision-maker for clinical care
- Patient consent for AI use
- TGA framework for medical device AI
- Documentation and clinical accountability
Legal
- Citation verification non-negotiable
- Confidentiality and privilege
- Court expectations
- Professional indemnity considerations
Education
- Academic integrity
- Disclosure of AI use
- Maintaining genuine learning
- School/university policies
Finance and tax
- Verify with ATO/professional bodies
- AFSL/TASA considerations
- Don’t provide unauthorised advice
- Honest dealings with clients
Journalism
- Verify every fact
- Source confidentiality
- Quote integrity
- Disclosure expectations
Government
- Agency frameworks
- Public trust
- Procedural fairness
- Records management
HR and recruitment
- Anti-discrimination laws
- Privacy of candidates
- Fair Work Act
- Human oversight of decisions
Research
- Research ethics
- Honest representation of methods
- Reproducibility
- Citation and attribution
The “would I be comfortable” test
For any AI use, ask:
“Would I be comfortable if the people affected knew everything about how I’m using AI?”
If yes: proceed. If no: reconsider.
Variations:
- “Would my client be comfortable?”
- “Would my regulator be comfortable?”
- “Would my employer be comfortable?”
- “Would my family be comfortable?”
- “Would the affected community be comfortable?”
When AI is the wrong tool entirely
For these situations, AI is the WRONG tool:
- Crisis support → human services (Lifeline, Beyond Blue, etc.)
- Medical emergencies → 000 or hospital
- Mental health emergencies → professional services
- Legal emergencies → lawyer
- Specific clinical decisions → qualified clinician
- Specific legal advice → qualified lawyer
- Specific financial advice → licensed financial adviser
- Replacing human connection → other humans
- Major life decisions without your judgment → your judgment + advisors
The honesty principles
To yourself
- Honest about AI’s role in your work
- Honest about your understanding
- Honest about what you’ve verified
To others
- Honest about AI use when expected
- Honest in content you produce
- Honest in attributions
About AI
- Honest about its capabilities (and limitations)
- Honest about its errors
- Honest about your dependence
Don’t be fooled
Don’t believe AI is “intelligent” the way humans are
- AI is sophisticated pattern matching
- AI doesn’t understand the way humans understand
- AI’s confidence isn’t knowledge
Don’t trust AI confidence
- Confidence isn’t accuracy
- Plausibility isn’t truth
- Verify important claims
Don’t anthropomorphise inappropriately
- AI doesn’t care about you
- AI doesn’t remember you (without explicit memory)
- AI is software, not friend
Don’t assume neutrality
- AI training data has biases
- AI outputs reflect training
- Bias awareness matters
Context-specific quick rules
Working with children
- Adult supervision
- Privacy heightened
- Age-appropriate content
- School policies
Working with vulnerable populations
- Heightened consent
- Cultural sensitivity
- Trauma-informed approach
- Specialist guidance
Indigenous Australians
- AIATSIS data sovereignty
- Cultural protocols
- Indigenous-led where relevant
- Engage Indigenous voices
Multicultural Australia
- Cultural humility
- Translation accuracy
- Community engagement
- Avoid stereotyping
Disability community
- Accessibility considered
- “Nothing about us without us”
- Disabled people in development
- Don’t make assumptions
LGBTIQ+ community
- Inclusive language
- Avoid stereotypes
- Specific community knowledge
When in doubt
When facing an ethical AI question:
- Pause — don’t just proceed
- Consider whose interests are affected
- Apply principles above
- Seek advice from colleagues, professional bodies, ethics resources
- Document your reasoning
- Choose the more conservative path if genuinely unsure
It’s better to under-use AI ethically than over-use AI unethically.
Australian-specific resources
Government
- Australian Government AI Ethics Principles
- Office of the Australian Information Commissioner (OAIC)
- Australian Human Rights Commission
- Australian Communications and Media Authority (ACMA)
Industry bodies
- Most have ethical guidance for AI use
Education
- Each university has academic integrity policy
Health
- AHPRA and professional boards
Cultural
- AIATSIS for Indigenous
- Various multicultural bodies
The bottom line
AI is a powerful tool. With power comes responsibility. The technology helps; the ethics are yours.
- Use AI to amplify your good work
- Don’t use AI to shortcut professional responsibilities
- Be honest about AI use
- Verify what matters
- Respect people’s autonomy and consent
- Maintain your accountability
- Stay aware of impacts beyond yourself
See also
- ai-safety-cheat-sheet — practical safety
- ai-prompting-cheat-sheet — better use
- ai-safety-primer — deeper context
- hallucinations — verification context
- australian-privacy-considerations
- fact-check-ai-output — verification
- All decision frameworks for specific contexts
Sources
- Australian Government AI Ethics Principles (2019, updated)
- Various professional body ethics codes
- Personal experience navigating AI ethics
- AI ethics research community
- Practical wisdom of Australian professional ethics traditions