Responsible AI Use
Knowing how to use AI responsibly in your own work—as a developer, student, professional, or consumer—is as important as understanding how AI is built. Responsible use protects you, your organisation, and the people affected by your decisions.
Responsible Use Principles
Verify before you trust: AI outputs are not automatically authoritative. Verify claims, especially in high-stakes contexts, against credible sources.
Be transparent: if you used AI to help write a document, generate code, or produce an image, disclose it where expected by your professional, academic, or platform norms.
Do not automate harm: using AI to generate phishing emails, deepfake content, disinformation, or targeted harassment content is unethical and increasingly illegal.
Protect sensitive data: do not paste confidential personal data, proprietary business information, or sensitive credentials into public AI tools. Assume inputs to third-party AI services may be logged and used for training.
Deep Learning ⊂ Machine Learning ⊂ Artificial Intelligence
Responsible Use Checklist
- Am I using AI for a task where the stakes of an error are acceptable given the oversight I am providing?
- Have I verified factual claims the AI made before acting on them or publishing them?
- Am I disclosing AI use where disclosure is expected or required?
- Am I protecting confidential data by not pasting it into third-party AI tools?
- Is my use of AI replacing a human unfairly rather than assisting one appropriately?
- Would I be comfortable if the people affected by my use of AI knew about it?
Responsible Use in Code
# Responsible AI use guidelines as code
def check_responsible_use(use_case, involves_personal_data, stakes_level, human_review):
"""
A simple decision aid for responsible AI use.
stakes_level: 'low', 'medium', 'high'
human_review: bool - whether a human will review before action is taken
"""
issues = []
recommendations = []
if involves_personal_data:
issues.append("Uses personal data")
recommendations.append("Verify data minimisation and legal basis for processing")
if stakes_level == "high" and not human_review:
issues.append("High-stakes decision with no human review")
recommendations.append("Add mandatory human review before any action is taken")
if stakes_level == "high":
recommendations.append("Document the decision process and who is accountable")
recommendations.append("Create an appeals process for affected parties")
if not issues:
print(f"Use case '{use_case}': Appears responsible. Proceed with care.")
else:
print(f"Use case '{use_case}': {len(issues)} issue(s) found.")
for issue in issues:
print(f" Issue: {issue}")
print()
print(" Recommendations:")
for rec in recommendations:
print(f" - {rec}")
# Examples
check_responsible_use("Resume screening", True, "high", False)
print()
check_responsible_use("Autocomplete suggestions", False, "low", False)
print()
check_responsible_use("Medical diagnosis support", True, "high", True)Key Takeaways
- Knowing how to use AI responsibly in your own work—as a developer, student, professional, or consumer—is as important as understanding how AI is built.
- Am I using AI for a task where the stakes of an error are acceptable given the oversight I am providing?
- Have I verified factual claims the AI made before acting on them or publishing them?
- Am I disclosing AI use where disclosure is expected or required?