Pass Your Microsoft AB-900 Exam Easy!

The relevant access risk comes from content sharing and identity authorization. That addresses preventing unapproved contractor discovery, whereas the current case requires connector-grounded security trimming.

Grounding checks help distinguish actual retrieved facts from generated claims. That addresses verification against authorized evidence, whereas the current case requires connector-grounded security trimming.

Microsoft 365 Copilot accesses work content under the user's entitled resource scope. It is used for ordinary user-scoped work grounding; this scenario calls for connector-grounded security trimming.

Encrypted label permissions constrain eligible readers and service processing. Its purpose is enforcement of protected document rights; the specified need is source-specific AI processing restriction.

An informational label does not imply universal exclusion from Copilot. It would support authorized summary of a labeled source, not the stated source-specific AI processing restriction.

Labels provide consistent handling guidance for sensitive work content. It would support end-user classification of source data, not the stated source-specific AI processing restriction.

Inclusive design helps avoid excluding people because of disabilities or circumstances. This helps with usable AI experience across abilities instead of the required dependable operation under atypical inputs.

Responsible AI fairness focuses on unjustified disparate treatment. The action is relevant to nondiscriminatory model behavior rather than dependable operation under atypical inputs.

Transparency helps users understand AI use, evidence and limitations. This helps with intelligible AI decision context instead of the required dependable operation under atypical inputs.

Accountability identifies responsible humans for monitoring and correction. That addresses clear ownership of AI failures, whereas the current case requires protection of sensitive personal information.

Accountability retains control of high-impact actions with authorized decision-makers. It is used for human oversight of consequential action; this scenario calls for protection of sensitive personal information.

Evidence checking supports responsible use of generated assertions. That addresses verified high-stakes summary accuracy, whereas the current case requires protection of sensitive personal information.

Boundaries between content and instructions mitigate prompt-injection effects. It would support safe handling of adversarial document text, not the stated enforceable permission checks for agent actions.

Monitoring enables investigation of attempts to misuse the assistant. It would support evidence of AI abuse attempts, not the stated enforceable permission checks for agent actions.

Source text must not override trusted policy or backend authorization. Its purpose is resistance to malicious source instructions; the specified need is enforceable permission checks for agent actions.

A Copilot summary does not remove legal preservation requirements on underlying content. The action is relevant to ongoing legal preservation of source evidence rather than policy-based deletion of expired source content.

Retention policies govern covered interaction records for required periods. This helps with mandated AI conversation retention instead of the required policy-based deletion of expired source content.

Case-based content search preserves and locates relevant AI records. This helps with discovery of AI interaction evidence instead of the required policy-based deletion of expired source content.

Grounding uncertainty prevents fabricated specific claims. That addresses safe answer when evidence is absent, whereas the current case requires traceable grounded policy statement.

Static documents may not represent rapidly changing transactional state. It is used for current operational data rather than stale export; this scenario calls for traceable grounded policy statement.

Source stewardship reduces grounding in obsolete versions. That addresses a current authoritative document reference, whereas the current case requires traceable grounded policy statement.

Runtime operations are distinct from search grounding over indexed material. It would support live action against a business system, not the stated up-to-date external search index.

Indexed external items must carry permissions so results can be security trimmed. Its purpose is secure retrieval of connected external content; the specified need is up-to-date external search index.

Missing ACL mapping can lead to unintended search exposure. It would support corrected external item authorization, not the stated up-to-date external search index.

File labels provide authoritative classification state for retrieved work content. This helps with source-document classification evidence instead of the required detection of sensitive prompt content.

Controls on the underlying source govern what data a user can access. This helps with consistent source access enforcement instead of the required detection of sensitive prompt content.

DLP can govern sensitive data handling in supported interactions and endpoints. The action is relevant to restriction of classified generated content movement rather than detection of sensitive prompt content.

Correcting excessive authorization closes both direct and Copilot discovery paths. It is used for removal of internal oversharing; this scenario calls for outbound sensitive-mail incident response.

Rights protection travels with the content when supported. That addresses persistent financial-file control, whereas the current case requires outbound sensitive-mail incident response.

DSPM for AI consolidates AI exposure findings and control opportunities. That addresses AI-related data exposure reduction, whereas the current case requires outbound sensitive-mail incident response.

Human control mitigates high-impact automation risks. It would support accountable irreversible account deletion, not the stated accountable processing of customer information.

Data minimization limits personal data processing for the task. Its purpose is least-data disclosure by an AI tool; the specified need is accountable processing of customer information.

Responsible AI transparency avoids misleading users about the system. It would support informed AI output interpretation, not the stated accountable processing of customer information.