Portfolioβ€ΊAI & Promptingβ€ΊGuardrails & Safety Prompting
Topic

Guardrails & Safety Prompting

Write prompts that actively prevent prompt injection, jailbreaks, and off-topic responses. Tests adversarial thinking and AI safety.

Adversarial testingPrompt injection defenceTopic restrictionAI safety

Choose Your Level

Pick the difficulty that matches where you are. You can come back and try a harder level later.

Topic Execution Guide

AI Safety, Content Filtering & Output Guardrails

Safety guardrails protect enterprise applications from unsafe content, topic drift, and privacy violations. AI safety developers implement NeMo Guardrails or Guardrails AI, PII detection masking, toxicity filters, and topic boundary enforcement.

1. AI Safety Guardrails Policy Specification

Policy document establishing input/output guardrail rules, toxicity thresholds, and banned topic boundaries.

2. PII & Toxicity Masking Pipeline Spec

Regex and NER pipeline detecting and redacting PII (emails, SSNs, credit cards) before sending queries to LLM APIs.

3. Fallback Response & Boundary Test Log

Security test log verifying graceful fallback responses when users attempt off-topic or unsafe queries.

Frequently Asked Questions (Guardrails & Safety Prompting)

What is the difference between Input Guardrails and Output Guardrails?

Input Guardrails inspect user queries BEFORE reaching the LLM to block malicious prompts or PII. Output Guardrails inspect LLM responses BEFORE rendering to users to catch hallucinations, toxicity, or data leaks.

How do topic boundary guardrails work?

Topic guardrails use semantic intent classifiers to verify user queries match approved application domains (e.g. banking support), returning polite fallback messages for off-topic requests.

Why is PII masking necessary before sending prompts to external APIs?

Masking PII prevents sensitive customer data from being transmitted to third-party cloud LLM providers, ensuring compliance with POPIA and GDPR.

Explore AI & Prompting Career Paths

Build proof of work across other topics or view full career roadmaps mapping technical skills to hiring expectations.