AI Threat Alert
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.
| Severity | CVE | Headline | Package | CVSS |
|---|---|---|---|---|
| LOW | CVE-2025-4287 | PyTorch NCCL: local DoS in distributed training reduce op | 3.3 | |
| MEDIUM | CVE-2025-46152 | PyTorch: OOB write causes incorrect bitwise shift results | pytorch | 5.3 |
| MEDIUM | CVE-2025-46153 | PyTorch: Dropout inconsistency enables membership inference | pytorch | 5.3 |
| MEDIUM | CVE-2025-53621 | DSpace: XXE injection enables server file disclosure | 6.9 | |
| HIGH | CVE-2024-34072 | SageMaker SDK: pickle deserialization enables RCE | 7.8 | |
| HIGH | CVE-2025-6921 | Transformers: ReDoS in optimizer halts training pipelines | transformers | 7.5 |
| CRITICAL | CVE-2025-62608 | mlx: security flaw enables exploitation | mlx | 9.1 |
| HIGH | CVE-2022-0736 | MLflow: insecure temp file handling causes DoS | mlflow | 7.5 |
| CRITICAL | CVE-2023-1177 | MLflow: path traversal allows arbitrary file read/write | mlflow | 9.8 |
| HIGH | CVE-2023-30172 | MLflow: path traversal exposes arbitrary server files | mlflow | 7.5 |
| CRITICAL | CVE-2023-2780 | MLflow: path traversal allows arbitrary file read/write | mlflow | 9.8 |
| CRITICAL | CVE-2023-3765 | MLflow: path traversal allows arbitrary file read | mlflow | 10.0 |
| HIGH | CVE-2023-4033 | MLflow: OS command injection enables local code execution | mlflow | 7.8 |
| HIGH | CVE-2023-6015 | MLflow: unauthenticated arbitrary file write via PUT | mlflow | 7.5 |
| CRITICAL | CVE-2023-6018 | MLflow: unauth file overwrite enables model poisoning | mlflow | 9.8 |
| CRITICAL | CVE-2023-6014 | MLflow: auth bypass allows arbitrary account creation | mlflow | 9.8 |
| HIGH | CVE-2023-43472 | MLflow: unauth REST API leaks sensitive ML data | mlflow | 7.5 |
| HIGH | CVE-2023-6709 | MLflow: SSTI enables RCE in ML experiment tracking | mlflow | 8.8 |
| HIGH | CVE-2023-6753 | MLflow: path traversal exposes arbitrary file read/write | mlflow | 8.8 |
| HIGH | CVE-2023-6831 | MLflow: path traversal allows arbitrary file write | mlflow | 8.1 |