Exam Bundle

This result would support retirement and disposal, a distinct governance need illustrated by: A discontinued predictive service remains callable with old credentials and retained training data after business shutdown. For the reported privacy/security policy integration issue at a multinational enterprise in Amsterdam, the relevant corrective action is: Extend existing privacy, security, and data-governance controls to AI inputs, logs, outputs, and service integrations. The offered observation does not establish whether that targeted control succeeded.

This result would support ai acceptable-use policy, a distinct governance need illustrated by: Employees paste confidential documents into unapproved public AI tools despite an existing generic IT policy. For the reported privacy/security policy integration issue at a multinational enterprise in Amsterdam, the relevant corrective action is: Extend existing privacy, security, and data-governance controls to AI inputs, logs, outputs, and service integrations. The offered observation does not establish whether that targeted control succeeded.

This result would support documentation and traceability, a distinct governance need illustrated by: A team cannot reconstruct which model version, policy, dataset, or approval produced a disputed AI decision. For the reported privacy/security policy integration issue at a multinational enterprise in Amsterdam, the relevant corrective action is: Extend existing privacy, security, and data-governance controls to AI inputs, logs, outputs, and service integrations. The offered observation does not establish whether that targeted control succeeded.

This response relates instead to insurance underwriting fairness, for which an example concern is: An insurer approves a pricing model without considering whether sensitive proxies violate applicable insurance discrimination rules. It could be useful in that different situation, but it would not directly address the legal regime inventory deficiency reported at a housing services organization in Manchester. The targeted acceptance criterion here is: The compliance register identifies overlapping duties and assigns owners for each applicable law.

This response relates instead to credit fairness, for which an example concern is: A lender checks statistical prediction accuracy but ignores how its automated score affects borrowers subject to fair-lending rules. It could be useful in that different situation, but it would not directly address the legal regime inventory deficiency reported at a housing services organization in Manchester. The targeted acceptance criterion here is: The compliance register identifies overlapping duties and assigns owners for each applicable law.

This response relates instead to product safety liability, for which an example concern is: A manufacturer integrates AI into a safety device but evaluates only the vendor's standalone model accuracy. It could be useful in that different situation, but it would not directly address the legal regime inventory deficiency reported at a housing services organization in Manchester. The targeted acceptance criterion here is: The compliance register identifies overlapping duties and assigns owners for each applicable law.

This result would support copyright in training data, a distinct governance need illustrated by: A model team scrapes copyrighted technical manuals for training and treats public accessibility as a universal license. For the reported legal regime inventory issue at a housing services organization in Manchester, the relevant corrective action is: Map all applicable legal regimes and obligations by market, data, product, and use case. The offered observation does not establish whether that targeted control succeeded.

This result would support confidential business material, a distinct governance need illustrated by: An employee uploads trade-secret pricing data to a third-party assistant without checking confidentiality commitments. For the reported legal regime inventory issue at a housing services organization in Manchester, the relevant corrective action is: Map all applicable legal regimes and obligations by market, data, product, and use case. The offered observation does not establish whether that targeted control succeeded.

This result would support credit fairness, a distinct governance need illustrated by: A lender checks statistical prediction accuracy but ignores how its automated score affects borrowers subject to fair-lending rules. For the reported legal regime inventory issue at a housing services organization in Manchester, the relevant corrective action is: Map all applicable legal regimes and obligations by market, data, product, and use case. The offered observation does not establish whether that targeted control succeeded.

This response relates instead to bias mitigation validation, for which an example concern is: A team changes training weights to address one group's error disparity without checking if performance worsened for others. It could be useful in that different situation, but it would not directly address the training experiment reproducibility deficiency reported at a professional services group in London. The targeted acceptance criterion here is: An authorized reviewer can reproduce the reported result or explain measured variance.

This response relates instead to class-imbalance testing, for which an example concern is: A rare safety incident makes up very little of a dataset, but the team celebrates a high overall accuracy score. It could be useful in that different situation, but it would not directly address the training experiment reproducibility deficiency reported at a professional services group in London. The targeted acceptance criterion here is: An authorized reviewer can reproduce the reported result or explain measured variance.

This response relates instead to adversarial robustness, for which an example concern is: An image model passes clean benchmarks but fails when small realistic perturbations are introduced. It could be useful in that different situation, but it would not directly address the training experiment reproducibility deficiency reported at a professional services group in London. The targeted acceptance criterion here is: An authorized reviewer can reproduce the reported result or explain measured variance.

This result would support bias mitigation validation, a distinct governance need illustrated by: A team changes training weights to address one group's error disparity without checking if performance worsened for others. For the reported training experiment reproducibility issue at a professional services group in London, the relevant corrective action is: Record experiment configuration, data and code versions, evaluation splits, and reproducibility evidence. The offered observation does not establish whether that targeted control succeeded.

This result would support integration and end-to-end tests, a distinct governance need illustrated by: A model passes unit tests but the live application passes the wrong input field and silently reverses classifications. For the reported training experiment reproducibility issue at a professional services group in London, the relevant corrective action is: Record experiment configuration, data and code versions, evaluation splits, and reproducibility evidence. The offered observation does not establish whether that targeted control succeeded.

This result would support adversarial robustness, a distinct governance need illustrated by: An image model passes clean benchmarks but fails when small realistic perturbations are introduced. For the reported training experiment reproducibility issue at a professional services group in London, the relevant corrective action is: Record experiment configuration, data and code versions, evaluation splits, and reproducibility evidence. The offered observation does not establish whether that targeted control succeeded.

This response relates instead to on-premises option, for which an example concern is: An organization assumes moving an AI model on-premises automatically makes it safe despite missing patching and monitoring capabilities. It could be useful in that different situation, but it would not directly address the multimodal capability decision deficiency reported at a regional technology consortium in Auckland. The targeted acceptance criterion here is: Representative text-and-image cases meet the defined task and safety criteria.

This response relates instead to open-source license review, for which an example concern is: A vendor proposes an open-weight model and assumes its availability means unrestricted commercial use and no governance duties. It could be useful in that different situation, but it would not directly address the multimodal capability decision deficiency reported at a regional technology consortium in Auckland. The targeted acceptance criterion here is: Representative text-and-image cases meet the defined task and safety criteria.

This response relates instead to deployment fit assessment, for which an example concern is: A company selects a chatbot product without checking the language, user population, performance, or data needed for its actual service. It could be useful in that different situation, but it would not directly address the multimodal capability decision deficiency reported at a regional technology consortium in Auckland. The targeted acceptance criterion here is: Representative text-and-image cases meet the defined task and safety criteria.

This result would support small versus large model, a distinct governance need illustrated by: An edge device must operate offline, but a team selects a large remote model based only on headline benchmark scores. For the reported multimodal capability decision issue at a regional technology consortium in Auckland, the relevant corrective action is: Evaluate modality requirements and select supported multimodal processing if justified. The offered observation does not establish whether that targeted control succeeded.

This result would support classic versus generative choice, a distinct governance need illustrated by: A team chooses an open-ended generative model to perform a simple constrained numerical classification without comparing simpler alternatives. For the reported multimodal capability decision issue at a regional technology consortium in Auckland, the relevant corrective action is: Evaluate modality requirements and select supported multimodal processing if justified. The offered observation does not establish whether that targeted control succeeded.

This result would support fine-tuning governance, a distinct governance need illustrated by: A foundation model is fine-tuned on confidential case files without analyzing memorization, access controls, or changed behavior. For the reported multimodal capability decision issue at a regional technology consortium in Auckland, the relevant corrective action is: Evaluate modality requirements and select supported multimodal processing if justified. The offered observation does not establish whether that targeted control succeeded.

This response relates instead to algorithmic unfairness, for which an example concern is: An AI screening tool achieves high overall accuracy while a protected demographic group has substantially worse false-negative rates. It could be useful in that different situation, but it would not directly address the supervised learning limits deficiency reported at a regional service provider in Singapore. The targeted acceptance criterion here is: A data-readiness assessment confirms labels and suitable held-out evaluation data.

This response relates instead to safety versus raw accuracy, for which an example concern is: An industrial anomaly detector improves average accuracy but produces rare unsafe false negatives in a safety-critical process. It could be useful in that different situation, but it would not directly address the supervised learning limits deficiency reported at a regional service provider in Singapore. The targeted acceptance criterion here is: A data-readiness assessment confirms labels and suitable held-out evaluation data.

This response relates instead to autonomous agent boundaries, for which an example concern is: An agent is permitted to make irreversible account changes on its own because it performed well in a few chat demonstrations. It could be useful in that different situation, but it would not directly address the supervised learning limits deficiency reported at a regional service provider in Singapore. The targeted acceptance criterion here is: A data-readiness assessment confirms labels and suitable held-out evaluation data.

This result would support generative versus predictive scope, a distinct governance need illustrated by: A procurement team evaluates an open-ended text drafting model solely with a conventional numeric regression error metric. For the reported supervised learning limits issue at a regional service provider in Singapore, the relevant corrective action is: Assess whether labeled, representative examples exist before selecting supervised learning for the intended outcome. The offered observation does not establish whether that targeted control succeeded.

This result would support autonomous agent boundaries, a distinct governance need illustrated by: An agent is permitted to make irreversible account changes on its own because it performed well in a few chat demonstrations. For the reported supervised learning limits issue at a regional service provider in Singapore, the relevant corrective action is: Assess whether labeled, representative examples exist before selecting supervised learning for the intended outcome. The offered observation does not establish whether that targeted control succeeded.

This result would support algorithmic unfairness, a distinct governance need illustrated by: An AI screening tool achieves high overall accuracy while a protected demographic group has substantially worse false-negative rates. For the reported supervised learning limits issue at a regional service provider in Singapore, the relevant corrective action is: Assess whether labeled, representative examples exist before selecting supervised learning for the intended outcome. The offered observation does not establish whether that targeted control succeeded.

This response relates instead to confidential business material, for which an example concern is: An employee uploads trade-secret pricing data to a third-party assistant without checking confidentiality commitments. It could be useful in that different situation, but it would not directly address the consumer deception deficiency reported at a technology services group in Stockholm. The targeted acceptance criterion here is: Published consumer messages accurately state what was automated and what evidence exists.

This response relates instead to credit fairness, for which an example concern is: A lender checks statistical prediction accuracy but ignores how its automated score affects borrowers subject to fair-lending rules. It could be useful in that different situation, but it would not directly address the consumer deception deficiency reported at a technology services group in Stockholm. The targeted acceptance criterion here is: Published consumer messages accurately state what was automated and what evidence exists.

This response relates instead to copyright in training data, for which an example concern is: A model team scrapes copyrighted technical manuals for training and treats public accessibility as a universal license. It could be useful in that different situation, but it would not directly address the consumer deception deficiency reported at a technology services group in Stockholm. The targeted acceptance criterion here is: Published consumer messages accurately state what was automated and what evidence exists.

This result would support insurance underwriting fairness, a distinct governance need illustrated by: An insurer approves a pricing model without considering whether sensitive proxies violate applicable insurance discrimination rules. For the reported consumer deception issue at a technology services group in Stockholm, the relevant corrective action is: Require accurate marketing and output claims and prevent deceptive representations of AI capability. The offered observation does not establish whether that targeted control succeeded.

This result would support confidential business material, a distinct governance need illustrated by: An employee uploads trade-secret pricing data to a third-party assistant without checking confidentiality commitments. For the reported consumer deception issue at a technology services group in Stockholm, the relevant corrective action is: Require accurate marketing and output claims and prevent deceptive representations of AI capability. The offered observation does not establish whether that targeted control succeeded.

This result would support copyright in training data, a distinct governance need illustrated by: A model team scrapes copyrighted technical manuals for training and treats public accessibility as a universal license. For the reported consumer deception issue at a technology services group in Stockholm, the relevant corrective action is: Require accurate marketing and output claims and prevent deceptive representations of AI capability. The offered observation does not establish whether that targeted control succeeded.