What security issues should we consider when implementing generative AI?

For in‑house models, follow a zero‑trust approach with strict access controls. For service‑based models, ensure data encryption in transit and at rest. Maintain rigorous testing and human oversight to avoid code vulnerabilities and bias.