Threat Modeling Generative AI: What 11,658 Incidents and the Research Actually Show
Most real-world harm from generative AI does not come from prompt injection. Analysis of 11,658 incidents shows output handling and misinformation dominate.
Read article