A customer using a travel chatbot reports that the chatbot responds with the following:def choose_location(location):    if...

Deploying data loss prevention (DLP) is the best answer because the chatbot has exposed internal application logic rather than returning only the information intended for the customer. This represents a form of sensitive information disclosure , which CompTIA SecAI+ specifically identifies as an AI-system attack or failure condition. The objectives also emphasize data security controls such as classification, redaction, masking, and minimization to prevent inappropriate disclosure.

A DLP control can inspect generated responses for source code, confidential identifiers, credentials, internal business logic, or other restricted information and prevent that content from being returned externally. Retraining the model to eliminate bias is unrelated because no discriminatory or statistically skewed behavior is shown. Secure output handling is important when model-generated output is consumed by downstream applications and could lead to injection or execution vulnerabilities. Here, however, the immediate issue is disclosure of internal information directly to a user. Cookies also do not address the problem; they are associated with maintaining web sessions and are not a general mitigation for server-side request forgery. Therefore, DLP most directly addresses the exposed information.

================