For Individuals
- Learning & Certification
- Specialist Tracks & Tools
Consulting & Enterprise
- Programmes & Solutions
- Diagnostics & Evidence
Insights
- Analysis & Research
  - Executive Reports In-depth whitepapers for decision-makers
  - Academic Research Peer-reviewed publications and frameworks
- Editorial & Dispatches
  - The C-Suite AI Memo Weekly executive strategy newsletter
  - News & Commentary Press, updates, and market notes
Events
About
Contact
LOGIN

Skip to content

Safe AI for Developers: Shipping AI Features Without Shipping the Risk

The Safety Requirements That Land on Engineering Teams

Quiz: The Safety Requirements That Land on Engineering Teams

Threat Modelling an AI Feature Before You Write Code

Quiz: Threat Modelling an AI Feature Before You Write Code

Prompt Injection, Tool Abuse and Untrusted Content Handling

Quiz: Prompt Injection, Tool Abuse and Untrusted Content Handling

Secrets, Personal Data and What Leaves Your Trust Boundary

Quiz: Secrets, Personal Data and What Leaves Your Trust Boundary

Designing Evaluation Gates Into Continuous Integration

Quiz: Designing Evaluation Gates Into Continuous Integration

Writing Test Sets That Represent Real Users and Real Failures

Quiz: Writing Test Sets That Represent Real Users and Real Failures

Guardrails at Input, Output and Tool Call Level

Quiz: Guardrails at Input, Output and Tool Call Level

Structured Output, Schema Validation and Refusal Handling

Quiz: Structured Output, Schema Validation and Refusal Handling

Retrieval Design That Reduces Fabricated Answers

Quiz: Retrieval Design That Reduces Fabricated Answers

1 of 2

Previous Lesson

Quiz: Prompt Injection, Tool Abuse and Untrusted Content Handling

Safe AI for Developers: Shipping AI Features Without Shipping the Risk
Prompt Injection, Tool Abuse and Untrusted Content Handling
Quiz: Prompt Injection, Tool Abuse and Untrusted Content Handling

Time limit: 0

Quiz Summary

0 of 5 Questions completed

Questions:

Information

You have already completed the quiz before. Hence you can not start it again.

Quiz is loading…

You must sign in or sign up to start the quiz.

You must first complete the following:

Results

Quiz complete. Results are being recorded.

Results

0 of 5 Questions answered correctly

Your time:

Time has elapsed

You have reached 0 of 0 point(s), (0)

Earned Point(s): 0 of 0, (0)
0 Essay(s) Pending (Possible Point(s): 0)

Categories

Not categorized 0%

Review
Answered
Correct
Incorrect

Question 1 of 5

1. Question

What is a prompt injection attack?

A code vulnerability in the model's neural network
Correct Incorrect Correct answer

An attempt to inject instructions into model input to redirect its behaviour
Correct Incorrect Correct answer

A method to compress the model for faster inference
Correct Incorrect Correct answer

A technique to fine-tune the model on new data
Correct Incorrect Correct answer

Correct / 1 Points

Incorrect / 1 Points
Question 2 of 5

2. Question

Why do language models have no inherent defence against prompt injection?

Developers forgot to add the defence
Correct Incorrect Correct answer

Models are too slow to parse instructions
Correct Incorrect Correct answer

Models do not have formal syntax that distinguishes instructions from data
Correct Incorrect Correct answer

Prompt injection only affects open-source models
Correct Incorrect Correct answer

Correct / 1 Points

Incorrect / 1 Points
Question 3 of 5

3. Question

Which mitigation is most effective at preventing tool abuse where a model is tricked into invoking a tool improperly?

Using a more powerful model
Correct Incorrect Correct answer

Role-based access control that enforces permissions on the tool itself
Correct Incorrect Correct answer

Deleting the tool from the system
Correct Incorrect Correct answer

Asking users to be more careful
Correct Incorrect Correct answer

Correct / 1 Points

Incorrect / 1 Points
Question 4 of 5

4. Question

What is an indirect prompt injection attack?

A second prompt sent after the first one fails
Correct Incorrect Correct answer

An attack that hides instructions inside legitimate-looking data like documents or reviews
Correct Incorrect Correct answer

An attack that only affects indirectly trained models
Correct Incorrect Correct answer

A prompt that is longer than 2000 tokens
Correct Incorrect Correct answer

Correct / 1 Points

Incorrect / 1 Points
Question 5 of 5

5. Question

How does structured output schema validation help defend against prompt injection?

It makes the model smarter at understanding instructions
Correct Incorrect Correct answer

It constrains model output to a fixed format, preventing off-piste responses
Correct Incorrect Correct answer

It prevents users from submitting malicious input
Correct Incorrect Correct answer

It trains the model to reject bad instructions
Correct Incorrect Correct answer

Correct / 1 Points

Incorrect / 1 Points

Login

Accessing this course requires a login. Please enter your credentials below!

Username or Email Address

Password

Remember Me

Continue with Facebook

Continue with Google

Lost Your Password?

Register

Don't have an account? Register one!

Register an Account

Keep building practical AI skills

Short, practical courses from Tesseract Academy, founded by Dr Stelios Kampakis, CStat.

Explore Individual Learning

© 2026 Tesseract Academy. All Rights Reserved.