Skip to main content
Home/Risks/IBM2025/Data poisoning

Data poisoning

Sub-category
Risk Domain

Vulnerabilities that can be exploited in AI systems, software development toolchains, and hardware, resulting in unauthorized access, data and privacy breaches, or system manipulation causing unsafe outputs or behavior.

"A type of adversarial attack where an adversary or malicious insider injects intentionally corrupted, false, misleading, or incorrect samples into the training or fine-tuning datasets."

Supporting Evidence (1)

1.
"Poisoning data can make the model sensitive to a malicious data pattern and produce the adversary’s desired output. It can create a security risk where adversaries can force model behavior for their own benefit."

Other risks from IBM2025 (63)