ENT-00001967 · Benchmark

CyberSecEval

CyberSecEval is a AI benchmark associated with Meta, classified in LXKeys.world as AI Benchmark / Evaluation.

Level IV ConnectedActiveDocumented

Human record

Overview

CyberSecEval is a documented AI dataset or benchmark associated with Meta. The record identifies its role in evaluation, training, measurement or comparison of intelligent systems, with emphasis on the source context and the type of capability it helps assess.

Timeline

2023Initial benchmark release
CyberSecEval entered the documented public record in 2023. This event is retained at the precision supported by the Entry’s reviewed source history.

Capabilities

Documented source contextPublic reference anchorStructured classification

Known limitations

Benchmark scores depend on the exact dataset version, prompt or evaluation protocol, scoring implementation and contamination controls.Leaderboard performance should not be treated as a complete measure of real-world capability or safety.Benchmark relevance can decline as models, data and evaluation practices evolve.

Technical and structured detail

Technical description

Structured LXKeys.world registry record for CyberSecEval. Entity type: Benchmark; classification: AI Benchmark / Evaluation; creator/organization context: Meta AI. Canonical source anchor: https://github.com/meta-llama/PurpleLlama/tree/main/CybersecurityBenchmarks. The record tracks source authority, first public appearance, timeline, capabilities, limitations, registry status and graph relationships. Automatic refresh is limited to sources explicitly classified as OFFICIAL and enabled for updates; documentary and research references remain non-authoritative unless reviewed.

Controlled tags

Documented EntityAI SystemRegistry Candidate

Relationship graph

CyberSecEval

Open in full graph

CyberSecEval

Published byOutgoing

Meta

CyberSecEval is associated with Meta through its documented source context.

Machine readable layer

Structured entry data for human tools, AI systems and machine clients.

JSON API Graph JSON

{
    "@context": [
        "https://schema.org",
        {
            "lxw": "https://lxkeys.world/schema/"
        }
    ],
    "@type": "Thing",
    "identifier": "ENT-00001967",
    "name": "CyberSecEval",
    "alternateName": [],
    "additionalType": {
        "category": "Data and Evaluation",
        "type": "Benchmark",
        "subtype": "",
        "lxkeysEntity": false
    },
    "description": "CyberSecEval is a AI benchmark associated with Meta, classified in LXKeys.world as AI Benchmark / Evaluation.",
    "creator": "Meta AI",
    "url": "https://lxkeys.world/entry.php?id=ENT-00001967&lang=en",
    "sameAs": "https://github.com/meta-llama/PurpleLlama/tree/main/CybersecurityBenchmarks",
    "image": "",
    "lxkeysWorld": {
        "worldId": "ENT-00001967",
        "kind": "Data and Evaluation",
        "type": "Benchmark",
        "subtype": "",
        "classification": "AI Benchmark / Evaluation",
        "organization": "Meta",
        "originContext": "Public AI and technical record",
        "firstPublicAppearance": "2023",
        "currentStatus": "Active",
        "documentationStatus": "Documented",
        "documentationIndex": {
            "total": 74,
            "documentation": 25,
            "evidence": 13,
            "structure": 25,
            "relationships": 11,
            "level": "Level IV Connected"
        },
        "lxCalendarium": {
            "start_date_utc": "2023-04-01",
            "created_utc": "2026-06-16T23:59:29+00:00",
            "created_dypclt": "D-0 Y-2 P-3 C-3 L-21 T-3",
            "updated_utc": "2026-06-16T23:59:29+00:00",
            "updated_dypclt": "D-0 Y-2 P-3 C-3 L-21 T-3"
        },
        "facts": [],
        "capabilities": [
            "Documented source context",
            "Public reference anchor",
            "Structured classification"
        ],
        "limitations": [
            "Benchmark scores depend on the exact dataset version, prompt or evaluation protocol, scoring implementation and contamination controls.",
            "Leaderboard performance should not be treated as a complete measure of real-world capability or safety.",
            "Benchmark relevance can decline as models, data and evaluation practices evolve."
        ],
        "tags": [
            "Documented Entity",
            "AI System",
            "Registry Candidate"
        ],
        "timeline": [
            {
                "date": "2023",
                "title": "Initial benchmark release",
                "description": "CyberSecEval entered the documented public record in 2023. This event is retained at the precision supported by the Entry’s reviewed source history.",
                "source_url": "https://github.com/meta-llama/PurpleLlama/tree/main/CybersecurityBenchmarks",
                "verification_status": "source_backed_curated_baseline"
            }
        ],
        "relationships": [
            {
                "target": "Meta",
                "type": "Published by",
                "description": "CyberSecEval is associated with Meta through its documented source context.",
                "evidence_level": "documentary"
            }
        ],
        "sources": [
            {
                "label": "CyberSecEval repository",
                "url": "https://github.com/meta-llama/PurpleLlama/tree/main/CybersecurityBenchmarks",
                "source_type": "Official Repository",
                "verification_status": "verified",
                "authority": "TRUSTED_PRIMARY",
                "role": "code_repository",
                "update_enabled": false,
                "authority_basis": "curated-corpus-refresh-2026-09-08"
            }
        ],
        "canonical": [],
        "imageMeta": []
    }
}

CyberSecEval

Overview

Timeline

Capabilities

Known limitations

Technical description

Controlled tags

CyberSecEval

Contribute to this record

Approved comments