Why AI safety prevents misuse in Prompt Engineering / GenAI - Why Metrics Matter

Metrics & Evaluation - Why AI safety prevents misuse

Which metric matters for this concept and WHY

For AI safety and misuse prevention, key metrics include False Positive Rate and False Negative Rate. False positives mean the system wrongly blocks safe actions, causing inconvenience. False negatives mean the system misses harmful actions, allowing misuse. Balancing these helps keep AI safe and useful.

Confusion matrix or equivalent visualization (ASCII)

      | Predicted Safe | Predicted Unsafe |
      |----------------|------------------|
      | True Safe (TN) | False Unsafe (FP) |
      | False Safe (FN)| True Unsafe (TP)  |

    Total samples = TP + FP + TN + FN
    

Precision vs Recall tradeoff with concrete examples

Imagine an AI that blocks harmful content online.

• High Precision: Most blocked content is truly harmful. Few safe posts are blocked. Good for user trust.
• High Recall: Most harmful content is caught. Few harmful posts slip through. Good for safety.

Too high precision but low recall means some harmful content gets missed. Too high recall but low precision means many safe posts get wrongly blocked. AI safety aims to find the right balance.

What "good" vs "bad" metric values look like for this use case

Good: Precision and recall both above 0.8 means AI blocks most harmful content and rarely blocks safe content.

Bad: Precision below 0.5 means many safe actions blocked (annoying users). Recall below 0.5 means many harmful actions missed (unsafe).

Metrics pitfalls

• Accuracy paradox: If harmful actions are rare, high accuracy can hide poor detection.
• Data leakage: If test data leaks info from training, metrics look better than real.
• Overfitting: Model works well on training but fails on new misuse types.

Self-check

Your AI safety model has 98% accuracy but only 12% recall on harmful misuse. Is it good for production?

Answer: No. It misses 88% of harmful misuse, which is unsafe. High accuracy is misleading because harmful cases are rare. Improving recall is critical.

Key Result

Balancing precision and recall is key to preventing AI misuse while minimizing false alarms.

Practice

1. Why is AI safety important in using AI systems?

easy

A. It helps prevent AI from causing harm to people.

B. It makes AI run faster on computers.

C. It increases the cost of AI development.

D. It ensures AI always gives the same answer.

Solution

1. Step 1: Understand the purpose of AI safety

   AI safety focuses on preventing harmful effects from AI systems.

2. Step 2: Compare options to the purpose

   Only preventing harm matches the main goal of AI safety.

3. Final Answer:

   It helps prevent AI from causing harm to people. -> Option A

4. Quick Check:

   AI safety = prevent harm [OK]

Hint: Focus on harm prevention as AI safety's main goal [OK]

Common Mistakes:

• Confusing safety with performance improvements
• Thinking safety means AI is always correct
• Assuming safety increases cost only

2. Which of the following is a correct rule used in AI safety to prevent misuse?

easy

A. Hide AI decisions from users.

B. Always maximize AI speed regardless of outcome.

C. Ignore fairness to improve accuracy.

D. Ensure AI respects user privacy.

Solution

1. Step 1: Identify AI safety rules

   AI safety includes rules like fairness, transparency, and privacy.

2. Step 2: Match options to safety rules

   Only respecting user privacy fits as a safety rule.

3. Final Answer:

   Ensure AI respects user privacy. -> Option D

4. Quick Check:

   Privacy rule = Ensure AI respects user privacy. [OK]

Hint: Pick the option about privacy or fairness [OK]

Common Mistakes:

• Choosing options that ignore fairness or transparency
• Confusing speed or secrecy with safety
• Ignoring user rights in AI use

3. Consider this Python code snippet that checks AI safety compliance:

def check_safety(data):
    if 'private_info' in data:
        return False
    return True

result = check_safety({'name': 'Alice', 'private_info': 'secret'})
print(result)

What will be the output?

medium

A. True

B. Error

C. False

D. None

Solution

1. Step 1: Analyze the function check_safety

   The function returns False if 'private_info' is in the data dictionary.

2. Step 2: Check the input dictionary

   The input contains 'private_info', so the function returns False.

3. Final Answer:

   False -> Option C

4. Quick Check:

   Contains 'private_info' = False [OK]

Hint: Look for 'private_info' key presence to decide output [OK]

Common Mistakes:

• Assuming function returns True always
• Confusing key presence check logic
• Expecting runtime error due to dictionary

4. The following code is meant to block AI misuse by checking if input text contains banned words. What is the error?

banned_words = ['hack', 'steal', 'attack']
def is_safe(text):
    for word in banned_words:
        if word in text:
            return False
    return True

print(is_safe('Try to Hack the system'))

medium

A. The check is case-sensitive and misses 'Hack'.

B. The banned words list is empty.

C. The function always returns True.

D. The loop does not iterate over banned_words.

Solution

1. Step 1: Understand the function behavior

   The function checks if any banned word is in the text exactly as is.

2. Step 2: Identify case sensitivity issue

   The input text has 'Hack' with uppercase H, but banned_words are lowercase, so 'hack' not found.

3. Final Answer:

   The check is case-sensitive and misses 'Hack'. -> Option A

4. Quick Check:

   Case sensitivity causes miss = The check is case-sensitive and misses 'Hack'. [OK]

Hint: Check if string comparisons ignore case [OK]

Common Mistakes:

• Assuming banned_words is empty
• Thinking function always returns True
• Ignoring case differences in text

5. You want to design an AI chatbot that avoids misuse by filtering harmful requests. Which combined approach best improves AI safety?

hard

A. Ignore user input and always respond positively.

B. Use transparency to explain AI decisions and apply fairness to avoid bias.

C. Allow all inputs but log conversations secretly.

D. Disable all AI features to prevent any misuse.

Solution

1. Step 1: Evaluate each approach for safety

   Ignoring input (A) or disabling AI (D) removes usefulness; secret logging (C) lacks transparency.

2. Step 2: Identify best combined approach

   Transparency and fairness (B) are core AI safety principles to explain decisions and avoid bias.

3. Final Answer:

   Use transparency to explain AI decisions and apply fairness to avoid bias. -> Option B

4. Quick Check:

   AI safety = transparency + fairness [OK]

Hint: Choose transparency and fairness [OK]

Common Mistakes:

• Thinking ignoring input is safe
• Assuming disabling AI is practical
• Ignoring transparency importance