AI Assurance DirectorySafety-critical AI platform
Llama Prompt Guard 2
Evidence strength: Emerging How this is scored
Meta Llama Prompt Guard 2 BERT classifiers (86M and 22M) that label a prompt string benign or malicious for prompt injection and jailbreaks. Drop-in replacement for Prompt Guard. Designed for Llama 4 Scout/Maverick and Llama 3. Llama 4 Community License.
Certification evidence
Analyst assessment: 1 NIST AI RMF function or subcategory — see which with Pro
Certification relevance. Searched vendor and publisher pages; no cited clause, function or article.
No cited mapping or literature quote yet. Analyst assessments, if any, stay on Pro.
Listing detail and source evidence (Pro)
Included on Pro, Team and Enterprise. Public summaries, tags and official links on this page stay free.
Analyst assessments (Pro)
Included on Pro, Team and Enterprise. Public summaries, tags and official links on this page stay free.
Procurement evidence pack (Pro) — print cited mappings and analyst notes. The print view is not on this page.
- Vendor / publisher
- Meta
- Type
- Commercial
- Official site
- https://www.llama.com/docs/model-cards-and-prompt-formats/prompt-guard/
- Industries
- Frameworks named on cited source
- None tagged. The cited pages do not name a framework we cover.
- Last verified
Listing is not an endorsement, certification or recommendation. Capabilities, availability and framework support can change. Confirm with the vendor and the relevant authority. See the disclaimer.