Data poisoning attacks target vulnerabilities in the training process of generative AI models. By injecting malicious or manipulated data into the training dataset, attackers can influence the model's behavior, leading to incorrect or harmful outputs. This type of attack is particularly concerning in finance, where compromised models could generate biased or fraudulent outputs that have significant real-world consequences.

A is incorrect: Phishing attacks involve tricking users into revealing sensitive information through deceptive communication and are unrelated to generative AI training.

B is incorrect: Input attacks manipulate the model during operation, not training.

D is incorrect: Deepfakes are a product of GenAI used maliciously, not attacks on the models.