Which of the following is the MOST critical key risk indicator (KRI) for an AI system?
Answer(s): B
Model drift - changes in data patterns or the model’s behavior over time - directly affects the reliability, fairness, and safety of an AI system. Monitoring the rate of drift is critical because it signals when the model may no longer perform as intended, posing operational, ethical, or regulatory risks.
How can an organization BEST protect itself from payment diversions caused by deepfake attacks impersonating management?
Resilient payment approval processes - such as multi-step verification, dual authorization, and out-of-band confirmation - provide strong safeguards against deepfake impersonation attacks. These controls reduce the risk of fraudulent payments by ensuring that no single communication or request, even if convincingly impersonated, can authorize a financial transaction.
Which of the following technologies can be used to manage deepfake risk?
Answer(s): D
Blockchain can help manage deepfake risk by providing a tamper-evident, verifiable record of media origin and modifications. By ensuring content authenticity and traceability, organizations can detect and prevent the misuse of deepfakes in critical communications and transactions.
An organization is facing a deepfake attack intended to manipulate stock prices. The organization's crisis communication plan has been activated. Which of the following is MOST important to include in the initial response?
Answer(s): C
During a crisis such as a deepfake attack affecting stock prices, the immediate priority is to communicate accurate, authoritative information to the public and stakeholders. A pre-approved statement ensures timely, consistent messaging that mitigates misinformation and reputational damage, stabilizing market and public response while further investigations proceed.
Which of the following BEST reduces the risk of exposing sensitive data through the output of large language models (LLMs) in applications?
Data sanitization removes or obfuscates sensitive information from datasets and model outputs. For LLM applications, this practice directly reduces the risk of inadvertently exposing confidential or personal data through generated responses, ensuring privacy and regulatory compliance.
Post your Comments and Discuss ISACA AAISM exam prep with other Community members: