AI FLUENCY · PHASE 1: AWARENESS
Module 5 — Capabilities and Limitations | Lesson 020
What Are AI Bias, Privacy, and Safety Risks?
Understand how AI can produce unfair outcomes, expose sensitive information, or create harm — and how responsible use can reduce these risks.
Watch the 54-second visual lesson, then explore the detailed explanations, examples, and safeguards below.
Watch on YouTube →Introduction
Artificial Intelligence can help us learn, create, solve problems, and accomplish tasks.
However, AI systems can also make unfair recommendations, mishandle sensitive information, or contribute to harmful decisions.
These risks remind us that being capable does not automatically make an AI system safe or appropriate for every situation.
In this final lesson of Module 5 — Capabilities and Limitations, we explore three important areas of responsible AI use:
BIAS · PRIVACY · SAFETY
Understanding these risks helps people use AI more thoughtfully and recognize when additional safeguards are necessary.
1. What Is AI Bias?
AI bias occurs when a system produces systematically distorted or unfair results.
Bias can arise from several sources, including unrepresentative training data, historical inequalities, data labeling, system design, and the way AI is used.
A system may perform differently for different groups or circumstances without a justified reason.
An AI system trained on historical hiring records may reproduce unfair patterns, potentially disadvantaging qualified job applicants.
Such a result does not mean the AI deliberately intended to discriminate.
It means the system's outputs may reflect or reinforce problematic patterns.
Why it matters: Unfair AI-assisted decisions can affect access to jobs, education, financial services, and other opportunities.
2. What Are AI Privacy Risks?
Privacy concerns how personal information is collected, used, shared, stored, and protected.
AI privacy risks arise when personal or confidential data is processed or exposed without appropriate authorization or safeguards.
Information that may require protection includes:
Names, addresses, contact details
Account details and financial records
Medical records and sensitive details
Employee and confidential business data
A manager copies employee performance records into an AI service without checking organizational policies or appropriate data protections.
Even if the manager only wants a summary, the action may create privacy or confidentiality risks.
Why it matters: AI can be useful for analyzing information, but individuals and organizations must control what data is shared and how it is processed.
3. What Are AI Safety Risks?
AI safety risks involve situations where the outputs or actions of an AI system may contribute to harm.
These harms may arise from incorrect information, unreliable recommendations, inappropriate automation, or insufficient safeguards.
An AI assistant provides an incorrect recommendation about an important financial or medical decision. If someone acts on it without verification, the consequences could be serious.
The level of risk depends on what the AI system does, the context in which it operates, and whether its outputs affect real-world decisions.
For example, a mistake in a creative writing exercise usually has different consequences from a mistake in a medical or employment decision.
Why it matters: AI systems should be tested, supervised, and appropriately restricted when errors could cause significant harm.
4. How Are Bias, Privacy, and Safety Different?
These three areas can overlap, but each focuses on a different type of concern.
| Risk | Main Concern | Example |
|---|---|---|
| Bias | Unfair or distorted outcomes | Unjustified differences in recruitment recommendations |
| Privacy | Inappropriate handling of information | Exposing confidential applicant records |
| Safety | Potential harm from outputs or actions | Unreliable automated decisions affecting people |
All three risks can appear within the same AI application.
Therefore, responsible AI use requires considering the entire context rather than examining one risk in isolation.
5. Everyday Example: AI-Assisted Hiring
Imagine a company using AI to help shortlist applicants for job interviews.
The AI reviews applications and recommends candidates for further consideration.
This may help recruiters organize large numbers of applications, but it can also introduce risks.
Bias
The AI may disadvantage qualified applicants because of unfair patterns in its data or design.
Privacy
Applicant information may be exposed, retained, or shared without appropriate controls.
Safety
Unreliable recommendations may negatively affect employment opportunities if accepted without review.
A responsible organization would evaluate the system before deployment, control access to applicant information, monitor performance, and provide meaningful human oversight.
Applicants should also have appropriate ways to raise concerns or challenge consequential errors.
6. How Can We Reduce AI Risks?
Responsible AI practices can help reduce the likelihood and severity of harm.
A simple approach is to consider fairness, privacy, accuracy, and human accountability.
A Practical Responsible AI Workflow
Evaluate whether AI results create unjustified differences or disadvantage people.
↓
Limit sensitive data sharing and use appropriate access and security controls.
↓
Test outputs against reliable evidence and relevant performance requirements.
↓
Assign responsibility, enable review, and intervene when outcomes are inappropriate.
Fairness → Privacy → Verification → Oversight
These safeguards should be tailored to the task and the consequences of possible failures.
They reduce risk but do not guarantee that every harmful outcome will be prevented.
7. Responsible AI: Your Personal Checklist
Before using AI for an important task, ask yourself these five questions:
1. Fairness: Could this output unfairly affect someone?
2. Privacy: Am I sharing personal or confidential information unnecessarily?
3. Accuracy: Have I verified important claims and recommendations?
4. Safety: What could happen if the AI is wrong?
5. Accountability: Who reviews the output and remains responsible for the decision?
The goal is not to avoid every possible use of AI.
It is to use AI with appropriate care, understanding, and judgment.
8. Why Does This Matter for AI Fluency?
AI Fluency involves more than knowing how to generate answers or complete tasks.
It also means recognizing the possible consequences of using AI.
For individuals, this includes protecting personal information and checking important outputs.
For educators and professionals, it includes considering how AI affects students, colleagues, customers, and other stakeholders.
For organizations, responsible use also requires governance, appropriate policies, testing, monitoring, and accountability.
AI can expand human possibilities, but the quality of its contribution depends on how responsibly we design, deploy, and use it.
Key Takeaway
AI capability creates possibilities. Responsible AI use protects people.
Recognizing bias, protecting privacy, reducing safety risks, and maintaining human accountability are essential parts of AI Fluency.
Think About It
Imagine an AI application helping you make a decision that affects another person.
Could the recommendation be unfair?
Does it involve sensitive information?
What could happen if the result is wrong?
Who should review the final decision?
Your challenge: Choose one AI application and identify one possible bias, privacy, or safety risk and a safeguard to reduce it.
What Have You Learned?
After completing Lesson 020, you should be able to:
- Explain AI bias and recognize examples of unfair outcomes.
- Identify privacy risks associated with personal and confidential data.
- Describe how AI outputs and actions can create safety risks.
- Recognize how bias, privacy, and safety concerns can overlap.
- Apply basic safeguards and explain the importance of human oversight.
Module 5 Complete!
You have now explored the capabilities and limitations of modern Artificial Intelligence.
Across four lessons, you have studied:
Lesson 017: What Can Today's AI Do Well?
Lesson 018: What Are AI Hallucinations?
Lesson 019: Why Can AI Be Wrong Even When It Sounds Confident?
Lesson 020: What Are AI Bias, Privacy, and Safety Risks?
Module Review Questions
- What are five important capabilities of modern AI?
- Why can language models generate invented information?
- Why is a confident-sounding answer not necessarily accurate?
- How can bias, privacy, and safety risks affect people?
- What responsibilities should humans retain when using AI?
You have now reached Lesson 020 of the AI Fluency Awareness curriculum.
COMING NEXT
Module 6 — AI and the Human World
Having explored AI's capabilities and limitations, we now turn to its relationship with human learning, work, and society.
Lesson 021: How Is AI Changing Learning and Education?
Lesson 022: How Is AI Changing Jobs and Work?
Lesson 023: Why Do AI Ethics, Governance, and Human Oversight Matter?
Lesson 024: What Is AI Fluency, and Why Does Everyone Need It?
Module 6 will complete the first phase of our journey: Awareness.
Further Learning Resources
NIST — AI Risk Management Framework
UNESCO — Recommendation on the Ethics of Artificial Intelligence
Your AI Fluency Journey
AI Fluency progresses through four connected phases.
Twenty foundational lessons have now been developed across five Awareness modules.
Four more lessons will complete the planned Awareness phase.
