What are the main LLM safety and alignment concerns?
mediumAnswer
- Hallucination (false but confident output), prompt injection (adversarial content in retrieved text hijacks the model), jailbreaks (bypassing safety instructions), data leakage (training data extraction, PII leakage), toxic or biased output, and misuse.
- Mitigations: system prompts, content filters, tool-use sandboxing, red teaming, output moderation, and clear provenance/citations.
Check yourself — multiple choice
- Prompt injection is only a risk in chat UIs
- Retrieved untrusted content can carry prompt injections
- Jailbreaks affect only open-source models
- Safety layers eliminate all risks
Untrusted retrieved data can hide prompts that hijack the model.
#safety#guardrails
Practise LLMs & GenAI
214 interview questions in this topic.