Injection attacks against AI agents commonly rely on malicious or manipulated prompts designed to alter system behavior or bypass safeguards. Implementing input filtering directly addresses this threat by detecting and blocking harmful inputs before they are processed by the AI system.

AAISM governance practices identify ownership and accountability as the most critical element in any centralized AI inventory. An AI inventory provides oversight by cataloging all AI assets within an organization, and assigning responsibility ensures that each system has clear governance, monitoring, and compliance coverage. While use cases, training data, and registries are valuable metadata, they do not guarantee accountability. Without defined ownership, no party is responsible for addressing risk, bias, or incidents. Therefore, the most important information to include is ownership and accountability details for each AI system.
References:
AAISM Exam Content Outline - AI Governance and Program Management (AI Inventories and Oversight) AI Security Management Study Guide - Ownership and Accountability Structures

AAISM materials emphasize that the most effective preventive safeguard is to ensure input sanitization.
Preventive controls stop malicious or malformed inputs from reaching the model in the first place, thereby reducing the likelihood of prompt injection, evasion, or poisoning at inference time. Model output monitoring is a detective control, not preventive. Penetration testing is an assessment technique rather than a safeguard.
Differential privacy protects data privacy but does not prevent adversarial input manipulation. Therefore, the most important preventive safeguard in a new AI product is robust input sanitization.
References:
AAISM Study Guide - AI Technologies and Controls (Preventive vs. Detective Safeguards) ISACA AI Security Management - Input Validation in AI Systems

AAISM materials make clear that the best safeguard against sensitive information being leaked through the outputs of LLMs is data sanitization. This involves filtering, redacting, or masking sensitive content before the model can use it, thereby preventing unintended disclosure in outputs. Encryption protects confidentiality in storage and transmission but does not stop output leaks. Adversarial testing helps identify vulnerabilities but does not prevent exposure by itself. Least privilege access restricts who can interact with the model but does not sanitize the content of its outputs. The control most directly tied to preventing leakage is implementing data sanitization techniques.
References:
AAISM Exam Content Outline - AI Technologies and Controls (Data Leakage Prevention) AI Security Management Study Guide - Sensitive Data Controls in Generative AI

AAISM directs that when harmful or biased behavior is observed in a production AI system, the organization should enter a formal incident/variance handling workflow that begins with root cause analysis (RCA) to identify the source of deviation (data drift, concept drift, feature leakage, pipeline changes, control failures) and determine proportionate risk treatments.