The safeguards map / 3.4 Guardrails
OPEA GenAIComps guardrails
Drafted with AI
Drafted with AI from the cited sources under the direction of UnGovr staff. UnGovr holds editorial responsibility for this page. How this site is made
Microservices in the OPEA GenAIComps repository that check large language model (LLM) inputs and outputs for unsafe content with Llama Guard or WildGuard and can reject what they flag.
Where it sits
3.4 Guardrails (primary): Runs Llama Guard or WildGuard as a microservice that classifies LLM inputs and outputs as unsafe or not, and rejects flagged input or output.
The Guardrails service enhances the security of LLM-based applications
captured
An input guardrail can either reject the input, halting further processing.
captured
An output guardrail can reject the output, preventing it from being returned to the user.
captured
Facts
- Concern
- safetysecurity
- Names the law
- no
- Subject
- AI system
- Used by
- those who make itthose who run it
- Runs
- at runtime
- Policy comes from
- vendor defaults
Handoffs
None recorded. A mapped entry with no handoff is here for its foundation home.
Collector's note
The foundation page states the sandbox stage at contribution and no later change. The categories come from the chosen model (Llama Guard or WildGuard), so the owner sets no rules of their own. The same repository also lists bias, hallucination, personally identifiable information (PII), prompt injection and toxicity detection microservices; this entry describes the general input and output guardrail service.
Every quote is verbatim from a stored capture of the page linked under it. Placement is our reading of the product's own documentation; no publisher has reviewed or endorsed it. Rows CC BY 4.0. Names, marks and logos belong to their owners; logos identify each product and are not part of the dataset.