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 6 of 15
Current
SeverityCVECVSS
CRITICALCVE-2024-271329.6
HIGHCVE-2024-14837.5
HIGHCVE-2024-15608.1
HIGHCVE-2024-15937.5
HIGHCVE-2024-15947.5
MEDIUMCVE-2024-42635.4
HIGHCVE-2024-370608.8
HIGHCVE-2024-370618.8
HIGHCVE-2024-05208.8
HIGHCVE-2024-29287.5
HIGHCVE-2024-271347.0
HIGHCVE-2024-88597.5
HIGHCVE-2025-14737.1
MEDIUMCVE-2025-14745.5
CRITICALCVE-2025-112009.8
HIGHCVE-2025-102797.0
MEDIUMCVE-2023-28004.7
HIGHCVE-2025-332138.8
HIGHCVE-2025-332337.8
CRITICALCVE-2024-528039.8

Page 6 of 15