What is the focus of 'Securing AI'?

• A. The security of the AI systems themselves, such as protecting against prompt injection.
• B. Securing training data only.
• C. The security of data center networks.
• D. Compliance with privacy laws.

Answer: (A)

Securing AI focuses on protecting the AI system itself—the model, its prompting interface, the data it processes, and the deployment environment—so its behavior remains trustworthy and resilient. This includes defending against prompt injection, where crafted inputs steer the model in unintended or unsafe ways, as well as other threats like tampering with the model, leaking information through responses, unauthorized access to the API, and ensuring the integrity of the inference and monitoring pipelines. In short, it’s about how the AI operates securely in real-world use, not just the data it was trained on or the surrounding infrastructure.

The other options miss this broader scope: securing training data alone addresses only the data used to teach the model; securing data center networks targets infrastructure, not the AI’s behavior; and privacy-law compliance deals with regulatory requirements rather than the security of the AI system itself.