AAISM Practice Questions

A.  

compromise the confidentiality of output data from the model

B.  

compromise the confidentiality of model input data

C.  

manipulate the behavior of the model during development

D.  

undermine the integrity of the AI system’s outputs

A.  

Failure to adequately assess AI risk

B.  

Inability to sufficiently identify shadow AI within the organization

C.  

Unwillingness of large AI companies to accept updated terms

D.  

Insufficient legal team experience with AI

A.  

Take the AI model offline

B.  

Adjust the hyperparameters of the AI model

C.  

Create an emergency change request to correct the issue

D.  

Return to an earlier phase in the AI life cycle

A.  

Supervised

B.  

Reinforcement

C.  

Unsupervised

D.  

Machine learning (ML)

A.  

Automating vulnerability identification

B.  

Developing a testing strategy including AI-specific threat modeling and adversarial attack simulations

C.  

Testing team competencies in IT threat mitigation

D.  

Using standard penetration testing methods

A.  

Enhancing monitoring and detection of model failures and anomalies

B.  

Implementing access controls to protect the AI system from unauthorized use

C.  

Testing the AI infrastructure failover mechanisms

D.  

Increasing the detail of AI solution backup and restoration processes

A.  

Robust access control techniques

B.  

Regular data reconciliation

C.  

Cryptographic hash functions

D.  

Diverse data sourcing strategies

A.  

Enhancing training data to increase variance

B.  

Monitoring the system for misuse cases

C.  

Fine-tuning the AI model to decrease hallucinations

D.  

Implementing a secondary AI system to confirm images

A.  

Recursive chunking

B.  

Automated output validation

C.  

Content enrichment

D.  

Human oversight

A.  

Regularly retraining the model using a diverse data set

B.  

Applying differential privacy and masking sensitive patterns in the training data

C.  

Incorporating adversarial training to expose and neutralize potential triggers

D.  

Monitoring model outputs and suspicious patterns to detect trigger activations