vulnerabilities

Generative AI models are susceptible to a kind of cyberattack called “data poisoning,” whereby malicious actors intentionally manipulate known source material to change the model’s understanding of an issue.