The AAISM study framework describes evasion attacks as attempts to manipulate or probe a trained model during inference by using crafted inputs that appear normal but cause the system to generate inconsistent or erroneous outputs. Contradictory results from nearly identical queries are a typical symptom of evasion, as the attacker is probing decision boundaries to find weaknesses. Poisoning attacks occur during training, not inference, while membership inference relates to exposing whether data was part of the training set, and model exfiltration involves extracting proprietary parameters or architecture. The clearest indication of contradictory outputs from similar queries therefore aligns directly with the definition of evasion attacks in AAISM materials.
References:
AAISM Study Guide - AI Technologies and Controls (Adversarial Machine Learning and Attack Types) ISACA AI Security Management - Inference-time Attack Scenarios

When AI systems make consequential decisions over sensitive data, AAISM requires explicit performance thresholds tied to decision quality--i.e., accuracy (and related error/false-rate limits) aligned to business risk appetite and regulatory expectations. Availability and latency are important service metrics, but decision integrity and error bounds are primary risk drivers in sensitive contexts. Establishing, monitoring, and enforcing minimum accuracy thresholds (with subgroup performance checks) is essential to reduce harm, ensure fairness/compliance, and support auditability.

AAISM stresses that the largest and most immediate risk for generative AI is unintentional data leakage by employees, making acceptable-use policies and staff training the most effective short-term risk mitigation.
Users must be instructed not to input sensitive data, understand usage restrictions, and follow governance requirements.
Relying solely on third-party audits (D) is insufficient. IP agreements (A) and opt-out clauses (B) are contractual protections, not operational safeguards.
References: AAISM Study Guide - Generative AI Security Risks; Policy and Awareness as Primary Controls.

The most direct preventative control against data poisoning is robust data validation/ingestion gating:
provenance checks, schema and constraint validation, anomaly/outlier screening, label consistency tests, and whitelist/blacklist source controls before data reaches training pipelines. Larger datasets (A) don't inherently prevent poisoning; monitoring (C) is detective; updating a foundation model (D) does not address tainted inputs entering the pipeline.
References: AI Security Management (AAISM) Body of Knowledge - Adversarial ML Threats and Training-Time Attacks; Secure Data Ingestion and Validation Controls. AAISM Study Guide - Poisoning Prevention: Provenance, Validation, and Sanitization Gates.

Validation focuses on confirming that the AI solution fulfills intended business needs and use cases, ensuring it performs as expected in real-world conditions before or during procurement decisions.