This is not a replacement for the official study guide. Please read the official study guide first — use this exam prep only to reinforce or practice your understanding.
Free AAISM exam practice questions with answers and explanations, organized by domain and part.
Domain 2: AI Risk Management
Q401 Which responsibility MOST clearly belongs to the AI deployer under the shared responsibility model?
- A. Regularly auditing AI systems and testing privacy and security configurations
- B. Building bias-reduced training data sets for the provider's base model, as part of the regular audit cycle
- C. Managing the provider's model weights and inference-time artifacts
- D. Producing the provider's SOC 2 and ISO certification reports
Answer: A
Deployers audit systems, test configurations, manage data governance and access control. Training data practices, weights and attestation reports are provider tasks.
Domain 2: AI Risk Management
Q402 A deployer wants to understand data sources and transformations used by its AI solution. Which practice is MOST relevant?
- A. Data provenance tracking as part of data governance
- B. Switching AI vendors every month to avoid dependence
- C. Removing logs to reduce storage costs for audit data
- D. Increasing monitor size for the data science team workstations
Answer: A
Data governance includes provenance tracking to understand sources and transformations. Monitor size, log removal and constant vendor changes do not improve traceability.
Domain 2: AI Risk Management
Q403 Which provider responsibility is MOST important for transparency to the deployer?
- A. Hiding model weights and training details from all customers
- B. Withholding incident information until a regulator asks for it
- C. Providing system and model documentation and incident notifications
- D. Avoiding compliance certifications to reduce cost of operation
Answer: C
Providers supply documentation, notifications and compliance evidence such as certifications. Hiding information undermines transparency and trust.
Domain 2: AI Risk Management
Q404 Which item is MOST typically part of supply chain risk management in the shared responsibility model?
- A. Employee birthday records kept by the human resources team
- B. Printer toner stock levels at each branch location
- C. Company parking allocations for staff and visitors
- D. SBOMs, due diligence, SLAs and compliance reports
Answer: D
Supply chain risk management includes SBOMs, due diligence, SLAs, compliance reports and contractual safeguards. Toner, parking and birthdays do not.
Domain 2: AI Risk Management
Q405 An organization connects a new AI service to a decades-old system without modernization. What is the GREATEST integration risk?
- A. Automatic compliance with regulations through inherited controls
- B. Reduced need for testing since the legacy system is already proven in production and its interfaces rarely change
- C. Faster processing in all cases because legacy platforms are mature
- D. Legacy systems may lack the security and interfaces needed, exposing data and creating failures
Answer: D
Integrating AI with legacy systems is an integration risk, because legacy platforms may be unsupported, unpatched or incompatible. It does not guarantee performance gains or compliance. Testing remains necessary.
Domain 2: AI Risk Management
Q406 A company's AI tool generates content resembling copyrighted works. Which risk does this BEST represent?
- A. A payroll error caused by an unreviewed automation script
- B. Network latency between regional offices and the data center
- C. Hardware failure in the servers hosting the inference service
- D. Intellectual property infringement
Answer: D
AI trained on copyrighted works can produce outputs raising IP infringement and ownership concerns. Hardware failure, latency and payroll errors are different risks.
Domain 2: AI Risk Management
Q407 Who owns AI-generated output is unclear in a vendor contract. What is the BEST action?
- A. Assume the company owns everything the system generates by default, since the company pays for the subscription
- B. Define IP ownership and usage rights for inputs and outputs contractually
- C. Let individual employees decide how to treat each output
- D. Ignore the issue until a dispute arises with the vendor
Answer: B
Contracts should define ownership and rights over inputs, outputs and models, because ownership of AI outputs is a contested legal area. Assuming ownership leaves exposure.
Domain 2: AI Risk Management
Q408 Why does AI software supply chain risk management PRIMARILY go beyond traditional approaches?
- A. It applies only to hardware components such as servers and GPUs, rather than to software, data and model dependencies
- B. It ignores data and models and focuses on code libraries alone
- C. It is unrelated to reputation, which is handled by marketing teams
- D. It must address entire systems, not just individual components, to protect brand and trust
Answer: D
Traditional supply chain risk management addresses individual components, whereas AI needs a view of the entire system, including data and models, to protect reputation and trust. It includes data and models.
Domain 2: AI Risk Management
Q409 Which pair of dimensions is MOST accurately part of the AI software supply chain in addition to people, processes and technology?
- A. Color and font used in the interface and documentation
- B. Season and weather conditions affecting data center cooling
- C. Data and model
- D. Language and accent models used by the voice assistants
Answer: C
The AI software supply chain includes people, processes, technology, data and model dimensions. Data covers pipelines and datasets, and model covers algorithms and parameters.
Domain 2: AI Risk Management
Q410 Which activity BEST supports ongoing monitoring of a vendor's AI solution?
- A. Collecting metrics only once at go-live and filing the results with the project documentation for later reference
- B. Setting thresholds, monitoring frequencies and escalation paths for key metrics
- C. Relying on the vendor's press releases for performance updates
- D. Disabling alerts to avoid distracting the operations team
Answer: B
Effective monitoring defines clear thresholds, frequencies and preestablished escalation paths, and watches prompt quality, relevance, bias, sentiment and toxicity. One-time collection or no alerts misses issues.