Challenge - 5 Problems

🎖️

Guardrail Mastery

Get all challenges correct to earn this badge!

Test your skills under time pressure!

🧠 conceptual

intermediate

2:00remaining

Why are guardrails important for AI agents?

Imagine an AI agent that can make decisions on its own. Why do we need guardrails to control it?

AGuardrails make the agent run faster by optimizing its code.

BGuardrails remove the need for any human oversight.

CGuardrails allow the agent to learn without any supervision.

DGuardrails help prevent the agent from taking harmful or unintended actions.

Attempts:

2 left

❓ model choice

intermediate

2:00remaining

Choosing guardrails for an AI agent

You want to design guardrails for an AI agent that interacts with users. Which approach best prevents the agent from sharing private user data?

ADisable all logging to protect user privacy.

BImplement strict data access controls and monitor outputs for sensitive information.

CTrain the agent without any data to avoid privacy issues.

DAllow the agent to access all data freely to improve learning speed.

Attempts:

2 left

❓ metrics

advanced

2:00remaining

Measuring effectiveness of guardrails

You have an AI agent with guardrails that limit risky actions. Which metric best shows if the guardrails are working?

ANumber of times the agent attempts forbidden actions during testing.

BTotal number of actions the agent takes overall.

CThe agent's training loss value after 10 epochs.

DThe speed at which the agent completes tasks.

Attempts:

2 left

🔧 debug

advanced

2:00remaining

Debugging a guardrail failure in an AI agent

An AI agent with guardrails sometimes performs unsafe actions. Which code snippet best fixes the guardrail check to stop this?

Agentic_ai

def check_action(action):
    # Guardrail: block actions labeled 'unsafe'
    if action == 'unsafe':
        return False
    return True

AChange the condition to if action == 'safe':

BChange the condition to if action != 'unsafe':

CChange the condition to if 'unsafe' in action:

DChange the condition to if action == 'unsafe': return True

Attempts:

2 left

🧠 conceptual

expert

3:00remaining

Why guardrails reduce AI agent disasters in real-world use

In real-world applications, why do guardrails significantly reduce the chance of AI agent disasters?

ABecause they limit the agent's actions to safe, tested behaviors and prevent unexpected harmful outcomes.

BBecause they make the agent learn faster and more efficiently.

CBecause they allow the agent to ignore human instructions when needed.

DBecause they remove all randomness from the agent's decisions.

Attempts:

2 left