AI FLUENCY · PHASE 2: APPRECIATION
Module 11 — AI Opportunities, Risks, and Responsibilities
What Are the Limitations and Risks of AI, and Why Should We Understand Them?
Understanding what AI cannot reliably do helps us recognize what could go wrong.
Limitation → Risk → Responsible Response
Introduction: Can AI Be Wrong?
Artificial Intelligence can generate explanations, summarize documents, analyze information, and support decisions.
Its responses can appear fluent, detailed, and convincing. But impressive presentation does not guarantee reliability.
AI systems may produce inaccurate statements, misunderstand context, lack current information, or generate inappropriate recommendations.
What happens when a convincing AI response is wrong—and someone relies on it?
This lesson explores two closely connected ideas: AI limitations and AI risks.
Understanding these ideas helps people make better decisions about when, where, and how AI should be used.
1. What Is an AI Limitation?
An AI limitation is a restriction or weakness in what an AI system can reliably know, understand, or perform.
For example, a system may:
- Generate an incorrect explanation.
- Misunderstand an ambiguous question.
- Lack access to current information.
- Overlook important context.
- Perform unreliably on unfamiliar tasks.
An AI system's limitations depend on the specific model, its available tools, the information provided, and the task.
What might this AI system not reliably know, understand, or do?
2. What Is an AI Risk?
An AI risk is the possibility that developing, deploying, or using AI could contribute to an undesirable outcome.
A limitation can create or increase a risk, but the two are not identical.
Everyday Example: Examination Registration
Limitation
A college AI assistant provides an outdated examination registration deadline.
Risk
A student relies on the incorrect date and misses registration.
The limitation concerns the system's information or performance.
The risk concerns possible harm arising from the use of that information.
Limitation: What might AI get wrong?
Risk: What could happen as a result?
3. Common Limitations of AI
Inaccuracy
AI may generate incorrect facts, references, quotations, or explanations.
Limited Context
AI may lack information about the user's circumstances or requirements.
Uncertainty
AI may provide definite-sounding answers even when evidence is limited.
Variable Reliability
Good performance on one task does not guarantee success on another.
What Are AI Hallucinations?
In generative AI, a hallucination is a generated output that is false, fabricated, or unsupported.
For example, an AI assistant may provide the title of a scientific article that does not actually exist.
Because the response looks credible, a person may accept it without checking.
A confident AI answer is not necessarily a correct AI answer.
4. Common Risks Associated with AI
Not every AI risk arises from a factual error. Some risks involve fairness, information security, privacy, or inappropriate dependence on technology.
⚖️ Bias and Unfairness
Outputs may reflect unfair or misleading patterns and affect groups differently.
🔒 Privacy
Personal or confidential information may be exposed or used inappropriately.
🛡️ Security
Connected AI systems may introduce unauthorized access or unintended actions.
🧠 Overreliance
People may accept AI-generated answers without independent evaluation.
📢 Misinformation
Incorrect or misleading AI content may be distributed widely.
🌍 Unequal Access
AI benefits and performance may not be equally available to everyone.
These risks differ in likelihood and severity across AI systems, users, and applications.
Recognizing a risk does not mean that harm is inevitable. It means the possibility should be examined.
5. One Example: An Incorrect Examination Deadline
Imagine a student asking a college AI assistant:
"What is the last date to register for my examination?"
The AI provides a date that was correct during the previous academic year.
The AI assistant does not have sufficiently accurate or current institutional information.
The student may miss registration because of an incorrect deadline.
Use authoritative institutional information, provide official source links, indicate uncertainty when necessary, and direct consequential questions to human staff.
This illustrates why the real-world importance of a task matters when deciding how to use AI.
6. The Limitation–Risk–Responsible Response Framework
We can examine an AI-supported activity through three questions.
STEP 1
LIMITATION
What might AI not reliably know, understand, or do?
↓
STEP 2
RISK
What might go wrong, and who could be affected?
↓
STEP 3
RESPONSIBLE RESPONSE
What verification, safeguards, or human oversight are needed?
This is a practical learning framework. It does not guarantee that every risk will be eliminated, and other risks may exist even when an obvious technical limitation is absent.
A Learning Example
Limitation: AI may incorrectly explain a scientific process.
Risk: A student may learn an inaccurate concept.
Responsible Response: Compare important explanations with reliable educational materials and ask for clarification.
7. Why Do Context and Consequences Matter?
The same type of AI error can have very different consequences.
| Situation | Possible AI Error | Why It Matters |
|---|---|---|
| Casual brainstorming | An unsuitable idea | Often relatively easy to reject |
| Educational explanation | Incorrect scientific statement | May cause misunderstanding |
| Examination information | Wrong registration deadline | May disrupt academic progress |
| Health-related guidance | Incorrect recommendation | Could create significant health consequences |
Before relying on an AI contribution, consider:
- How important is the task or decision?
- Who could be affected?
- How easily could an error be detected?
- Could a mistake be corrected or reversed?
- What level of human review is appropriate?
The level of AI oversight should reflect the potential consequences.
8. Connecting AI Limitations with the 4Ds
The four capabilities introduced in Module 10 help us address AI limitations.
Delegation
Avoid assigning unsuitable or unreliable tasks to AI.
Description
Communicate requirements, limitations, and relevant context.
Discernment
Examine claims, verify information, and identify errors.
Diligence
Apply safeguards and take responsibility for AI-assisted work.
AI Fluency involves knowing not only how to use AI, but also when to question it, verify it, or choose a different approach.
Lesson 042 — Key Takeaway
Recognize Limitations.
Understand Risks.
Respond Responsibly.
AI Fluency means understanding what AI can do, what it may get wrong, and how human judgment protects meaningful outcomes.
Quick Knowledge Check
Answer each question before revealing its explanation.
Question 1
What is an AI limitation?
A. A restriction on what AI can reliably know or do
B. Every problem that has happened because of AI
C. Proof that AI can never be useful
Reveal Answer
A. A limitation concerns a system's reliable capabilities.
Question 2
What is an AI risk?
A. The speed at which AI works
B. The possibility of an undesirable outcome
C. Any task involving AI
Reveal Answer
B. Risks concern potential undesirable consequences.
Question 3
Which situation illustrates an AI hallucination?
A. AI correctly summarizes a public document
B. AI generates a nonexistent scientific reference
C. A user asks AI to rewrite a paragraph
Reveal Answer
B. Fabricated references are an example of AI hallucination.
Question 4
A student misses registration because AI gave an outdated deadline. What is the risk?
A. The student may miss registration
B. The AI uses text to answer questions
C. The AI response appears confident
Reveal Answer
A. Missing registration is the possible real-world consequence.
Question 5
Which is a responsible response to important AI-generated information?
A. Assume confident answers are correct
B. Publish immediately
C. Verify relevant claims with authoritative sources
Reveal Answer
C. Important claims should be checked against reliable sources.
Question 6
Should every AI activity receive exactly the same level of human oversight?
A. Yes, regardless of the consequences
B. No, oversight should reflect the task and potential consequences
C. No, AI never needs evaluation
Reveal Answer
B. More consequential activities generally require stronger safeguards.
Think About It: A College AI Assistant
Your college introduces an AI assistant that answers questions about courses, examinations, and campus services.
How would you protect students from incorrect information?
Your Three-Part Reflection
1. Limitation
What might the assistant misunderstand or answer inaccurately?
2. Risk
What consequences could follow if students rely on those responses?
3. Responsible Response
What official sources, human review processes, and escalation arrangements would reduce the risk?
Prepare a short Limitation–Risk–Responsible Response Blueprint for one everyday AI application.
What Have You Learned?
After completing Lesson 042, you should be able to:
- Distinguish an AI limitation from an AI risk.
- Recognize inaccurate and fabricated AI outputs.
- Identify concerns involving bias, privacy, security, and overreliance.
- Explain why consequences differ across AI applications.
- Recognize the importance of appropriate human oversight.
- Apply the Limitation–Risk–Responsible Response framework.
- Connect responsible responses with the four Ds of AI Fluency.
PHASE 2 · APPRECIATION
Module 11 — Your Learning Journey
✓ Lesson 041
Recognizing AI Opportunities and Risks
● Lesson 042 — Current
Understanding AI Limitations and Risks
Lesson 043 — Coming Next
Responsible AI Use and Human Accountability
Lesson 044
Balancing AI Opportunities, Risks, and Responsibilities
COMING NEXT · LESSON 043
What Does Responsible AI Use Mean, and Who Is Accountable for Its Outcomes?
Understanding AI limitations helps us recognize the need for appropriate safeguards.
Lesson 043 examines human accountability, transparency, fairness, privacy, oversight, and responsible AI use.
If AI contributes to a decision, who remains responsible for its consequences?
Continue Your Learning
Previous Lesson — 041
How Can We Recognize AI Opportunities While Understanding Their Risks?
Review Module 10
The 4Ds: Delegation, Description, Discernment, and Diligence.
Phase 1 — Awareness
Explore Phase 1 Learning Hub →
AI Fluency Master Roadmap
Further Learning Resources
NIST — AI Risk Management Framework
Use AI with Understanding, Not Blind Trust.
Recognize what AI may not reliably do.
Consider what might go wrong.
Choose safeguards appropriate to the consequences.
Recognize Limitations.
Understand Risks.
Respond Responsibly.
