Rate limiting and abuse prevention in Prompt Engineering / GenAI - Cheat Sheet & Quick Revision

Recall & Review

beginner

What is rate limiting in the context of AI services?

Rate limiting is a technique used to control how many requests a user or system can make to an AI service in a given time. It helps prevent overload and abuse by limiting excessive use.

beginner

Why is abuse prevention important for AI models?

Abuse prevention protects AI models from harmful or excessive use that can cause service disruption, unfair resource use, or biased outputs. It ensures fair access and maintains system reliability.

intermediate

Name two common methods used for rate limiting.

Two common methods are:<br>1. Token Bucket: Users get tokens at a fixed rate; each request uses a token.<br>2. Fixed Window: Limits requests in fixed time windows, like 100 requests per minute.

intermediate

How can machine learning help in abuse prevention?

Machine learning can detect unusual patterns or behaviors that suggest abuse, like sudden spikes in requests or suspicious inputs, and trigger protective actions automatically.

beginner

What is a real-life example of rate limiting?

A website might allow only 5 login attempts per minute to stop hackers from guessing passwords. This is rate limiting to prevent abuse.

What does rate limiting primarily help prevent?

A Too many requests in a short time

B Slow internet connection

C Incorrect AI predictions

D Data storage overflow

Rate limiting controls the number of requests to avoid overload and abuse.

Which method uses tokens to control request rates?

A Random Sampling

B Fixed Window

C Sliding Window

D Token Bucket

Token Bucket gives tokens at a fixed rate; each request consumes one token.

Why is abuse prevention important for AI services?

A To reduce training time

B To increase AI model size

C To protect from harmful or excessive use

D To improve color contrast

Abuse prevention ensures fair use and protects AI services from misuse.

Which of these is NOT a sign of abuse detected by machine learning?

A Sudden spike in requests

B Consistent normal usage

C Suspicious input patterns

D Repeated failed attempts

Consistent normal usage is expected and not a sign of abuse.

A website limits login attempts to 5 per minute. This is an example of:

A Rate limiting

B Data augmentation

C Model training

D Feature extraction

Limiting login attempts is a practical example of rate limiting.

Explain what rate limiting is and why it is important in AI services.

Think about how many times you can ask a question before being told to wait.

Describe how machine learning can help detect abuse in AI systems.

Consider how a security guard notices strange behavior.