1Training-Data PoisoningTraining-data poisoning turns a model's source of learning into a place where attackers can plant errors, bias, or hidden behavior.Model & data attacks 2Model Theft, Extraction & InversionHow attackers copy model behavior, infer hidden data, and what defenders can do without breaking useful access.Model & data attacks 3Membership Inference & Data LeakageHow models reveal what they should not, why membership inference works, and what practical teams can do to reduce the risk.Model & data attacks 4The AI Supply ChainAI systems are assembled from data, code, models, compute, and people; knowing that chain is how you manage cost, risk, and trust.Model & data attacks 5Agent Identity & AuthenticationBefore an agent can be trusted with tools, data, or decisions, a system must know which agent it is and how that claim is proved.Model & data attacks