As enterprises accelerate AI adoption, a new category of threat is drawing attention from security leaders. Capacity CIO Shatabdi Sharma is quoted in this InformationWeek piece examining distillation attacks — a technique that extracts capabilities from advanced AI models to train cheaper, less secure alternatives. Sharma breaks down the mechanics, the model provenance questions CIOs should be asking vendors, and why anonymizing data is a practical first line of defense for enterprises relying on frontier models.