Your question is AI Security in Legacy Systems. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
What are the primary security considerations when integrating AI into legacy infrastructure?
Discuss how you would assess risks and design an end-to-end ML integration without assuming a specific vendor, product, industry, or deployment model. Cover identity and authorization, data protection, model and prompt security, legacy-system isolation, monitoring, incident response, and safe degradation. Explain how you would validate the design, manage training-serving skew and feature drift where applicable, and introduce controls without disrupting existing workloads.