Skip to main content
Home/Risks/Gipiškis2024/Fine-tuning related (Fine-tuning dataset poisoning)

Fine-tuning related (Fine-tuning dataset 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 deployer can poison the dataset used during the fine-tuning process [98] to induce specific, often malicious, behaviors in a model. This can be performed without having access to the model’s weights. This poisoning can be difficult to detect through direct inspection of the dataset, as the manipulations may be subtle and targeted."(p. 14)

Other risks from Gipiškis2024 (144)