CoolFace
Datasetpublic

fevziegeyurtsevenler/owasp-agentic-top10-2026-tr

OWASP Top 10 for Agentic Applications 2026 — Turkish edition (machine-readable) Machine-readable Turkish edition of the OWASP Top 10 for Agentic Applications (2026) (ASI01:2026–ASI10:2026), each risk mapped to verified MITRE ATLAS techniques and the OWASP LLM Top 10. OWASP Agentic Uygulamalar icin Ilk 10 (2026) risklerinin makine-okunur Turkce surumu. UNOFFICIAL community Turkish translation. The authoritative text is the English OWASP source: OWASP GenAI Security Project - Top… See the full description on the dataset page: https://huggingface.co/datasets/fevziegeyurtsevenler/owasp-agentic-top10-2026-tr.

sourceHugging Facecc-by-sa-4.0updated 1mo agoView on Hugging Face
1likes27downloads
Dataset Card

OWASP Top 10 for Agentic Applications 2026 — Turkish edition (machine-readable)

Machine-readable Turkish edition of the OWASP Top 10 for Agentic Applications (2026) (ASI01:2026–ASI10:2026), each risk mapped to verified MITRE ATLAS techniques and the OWASP LLM Top 10. OWASP Agentic Uygulamalar icin Ilk 10 (2026) risklerinin makine-okunur Turkce surumu.

UNOFFICIAL community Turkish translation. The authoritative text is the English OWASP source: OWASP GenAI Security Project - Top 10 for Agentic Applications 2026. Resmi olmayan topluluk cevirisidir; yetkili metin Ingilizce OWASP kaynagidir.

Dataset structure

One record per OWASP Agentic risk (10 rows). Fields:

fielddescription
idOfficial OWASP id, e.g. ASI01:2026
en_titleOfficial English risk title
tr_titleTurkish title (unofficial translation)
en_summaryCondensed English summary of the official entry
tr_summaryTurkish summary (unofficial translation)
example_en / example_trReal-world example incident (EN/TR)
atlas_techniquesList of {id, name, verified} MITRE ATLAS techniques
owasp_llm_top10_relatedRelated OWASP LLM Top 10 (2025) ids
source_url, snapshot_dateProvenance

Load

python
import json, urllib.request

# from a local clone
with open('data/agentic-top10-2026.json', encoding='utf-8') as fh:
    data = json.load(fh)

for r in data['risks']:
    atlas = ', '.join(t['id'] for t in r['atlas_techniques'])
    print(r['id'], '-', r['en_title'], '/', r['tr_title'])
    print('   ATLAS:', atlas)

With the 🤗 datasets library (flatten the JSON's risks array):

python
from datasets import load_dataset
ds = load_dataset('json', data_files='data/agentic-top10-2026.json', field='risks')
print(ds['train'][0]['id'])  # ASI01:2026

Risks

IDEN titleTR baslikMITRE ATLAS
ASI01:2026Agent Goal HijackAjan Hedef KacirmaAML.T0051, AML.T0051.001, AML.T0051.000, AML.T0054
ASI02:2026Tool Misuse and ExploitationArac Kotuye Kullanimi ve IstismariAML.T0053, AML.T0051
ASI03:2026Identity and Privilege AbuseKimlik ve Ayricalik IstismariAML.T0012, AML.T0055
ASI04:2026Agentic Supply Chain VulnerabilitiesAgentic Tedarik Zinciri ZafiyetleriAML.T0010, AML.T0010.001, AML.T0011.001, AML.T0058
ASI05:2026Unexpected Code ExecutionBeklenmeyen Kod YurutmeAML.T0050, AML.T0011, AML.T0011.000, AML.T0072
ASI06:2026Memory and Context PoisoningBellek ve Baglam ZehirlenmesiAML.T0070, AML.T0020, AML.T0019
ASI07:2026Insecure Inter-Agent CommunicationGuvensiz Ajanlar-Arasi IletisimAML.T0051.001, AML.T0053
ASI08:2026Cascading FailuresZincirleme ArizalarAML.T0031, AML.T0048
ASI09:2026Human-Agent Trust ExploitationInsan-Ajan Guveni IstismariAML.T0052, AML.T0052.000
ASI10:2026Rogue AgentsHain AjanlarAML.T0057, AML.T0024, AML.T0025, AML.T0048

Provenance & honesty

  • —Source taxonomy: OWASP GenAI Security Project, Top 10 for Agentic Applications 2026 (released 2025-12-09), licensed CC BY-SA 4.0. https://genai.owasp.org/resource/owasp-top-10-for-agentic-applications-for-2026/
  • —ATLAS techniques: MITRE ATLAS (mitre-atlas/atlas-data) v5.6.0 (https://raw.githubusercontent.com/mitre-atlas/atlas-data/main/dist/ATLAS.yaml). Every id is verified against the vendored data/atlas_techniques_ref.json.
  • —Translation status: UNOFFICIAL community Turkish translation. The authoritative text is the English OWASP source. / Resmi olmayan topluluk cevirisidir; yetkili metin Ingilizce OWASP kaynagidir.
  • —ATLAS mapping status: Community/analytical mapping. Each ATLAS id is verified to exist in the cited ATLAS distribution (v5.6.0), but the risk->technique association is this edition's analysis, not an official OWASP<->ATLAS mapping. Technique ids and top-level names are verbatim from ATLAS; sub-technique names use ATLAS's canonical 'Parent: Sub' display form.
  • —Not endorsed by OWASP or MITRE. Independent community edition.

Citation

bibtex
@misc{owasp_agentic_top10_2026_tr,
  title        = {OWASP Top 10 for Agentic Applications 2026 - Machine-readable Turkish edition},
  author       = {Yurtsevenler, Fevzi Ege},
  year         = {2026},
  note         = {Unofficial community Turkish translation; taxonomy (c) OWASP GenAI Security Project (CC BY-SA 4.0); ATLAS ids (c) MITRE ATLAS.},
  howpublished = {\url{https://genai.owasp.org/resource/owasp-top-10-for-agentic-applications-for-2026/}}
}

License

Data/text: CC BY-SA 4.0 (matches the OWASP source). Renderer code in the repo: Apache-2.0.

İlgili AltaySec kaynakları