AI Component

Training Data

Training data is both the model's most valuable input and its most underprotected one. Three problem classes dominate. First, poisoning: an attacker who can influence a public dataset, a web crawl, or a fine-tuning corpus can plant backdoors or biases that survive into the deployed model — BadNets-style attacks on image classifiers, trigger-phrase attacks on LLMs, and reward-hacking on RLHF datasets. Second, memorization and leakage: models can regurgitate verbatim training data, exposing PII and copyrighted content; this has driven the active New York Times v. OpenAI litigation and is a recurring GDPR concern. Third, provenance: when training data origins are unclear, downstream users inherit legal and security risk they can't assess. EU AI Act Article 10 (Data Governance) and ISO 42001 Annex A treat training-data quality as a controlled asset. Defenses: data lineage tracking, deduplication, PII scrubbing before training, and adversarial training against known trigger families.

287
Total CVEs
15
Pages
Page 14 of 15
Current
SeverityCVECVSS
CRITICALCVE-2026-112403-
HIGHCVE-2026-1124088.1
HIGHCVE-2026-1141358.8
LOWCVE-2026-1157493.3
HIGHCVE-2026-578458.8
HIGHCVE-2026-1129407.1
LOWCVE-2025-85563.7
MEDIUMCVE-2026-1186845.9
MEDIUMCVE-2026-1187606.5
HIGHCVE-2026-1195787.8
HIGHCVE-2026-1215617.7
UNKNOWNCVE-2026-118623-
UNKNOWNCVE-2026-118626-
UNKNOWNCVE-2026-118624-
HIGHCVE-2026-1027397.8
HIGHCVE-2026-1232798.8
MEDIUMCVE-2026-1232866.5
HIGHCVE-2026-1232877.5
MEDIUMCVE-2026-1232894.3
HIGHCVE-2026-955197.8

Page 14 of 15