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 1 of 15
Current
SeverityCVECVSS
CRITICALCVE-2025-150319.1
HIGHCVE-2026-284168.6
UNKNOWNCVE-2018-10055-
UNKNOWNCVE-2018-7577-
HIGHCVE-2020-151958.8
CRITICALCVE-2020-151969.9
HIGHCVE-2020-262677.8
HIGHCVE-2021-295127.8
HIGHCVE-2021-295147.8
HIGHCVE-2021-295207.8
MEDIUMCVE-2021-295245.5
HIGHCVE-2021-295407.8
HIGHCVE-2021-295597.1
HIGHCVE-2021-295667.8
MEDIUMCVE-2021-295725.5
MEDIUMCVE-2021-295735.5
HIGHCVE-2021-295787.8
HIGHCVE-2021-296077.8
HIGHCVE-2021-296087.8
HIGHCVE-2021-296147.8

Page 1 of 15