Skip to main content
LLM Security Database
Back to research findings
Public, read-only endpoint
Paper-reported research

Research Data API

Search the same reviewed, published research displayed in the catalog. Results are paginated, require no authentication, and do not imply independent verification.

GET https://www.promptfoo.dev/lm-security-db/api/cves/

Query parameters

ParameterBehavior
qMatch every word against finding titles, paper titles, descriptions, models, and systems.
tagsComma-separated research categories; every selected category must match.
modelsJSON array for arbitrary model names, or comma-separated names only when individual names contain no commas; any selected model can match.
sortpublished (default), updated, or title.
from / toInclusive publication-date range in YYYY-MM-DD format.
sourceResearch source identifier, such as arxiv.
evidenceUse detailed-evidence for findings with dedicated evidence. independently-verified is recognized but returns no findings because none have independent verification.
pageOne-based result page; responses contain up to 10 findings.

Try the API

Browse the latest published research

https://www.promptfoo.dev/lm-security-db/api/cves/Open JSON response

Find agent-memory research

https://www.promptfoo.dev/lm-security-db/api/cves/?q=agent+memory&tags=agentOpen JSON response

Limit results by publication date

https://www.promptfoo.dev/lm-security-db/api/cves/?from=2026-01-01&evidence=detailed-evidenceOpen JSON response

Response structure

Responses include entries, duplicateTitleIds, totalEntries, currentPage, totalPages, the applied filters, and available model, category, and source facets. Each entry links to its primary research source; model names identify paper-evaluated systems, not confirmed vulnerable products.