unified
UC-AI-20 — Prevent harmful, out-of-scope, hallucinated, and over-exposed AI outputs
Filter and shape every AI output before release: block or transform content that matches the system's harmful-output taxonomy, keep responses within the declared scope and capabilities, detect agent-specific high-risk outputs and route them to defined responses by severity, ground factual claims in cited sources and verify them to limit hallucination, withhold system prompts, internal data, and other over-exposed information, and sanitize outputs consumed by downstream systems so they cannot carry executable or injected payloads. Measure filter effectiveness and retain configurations, block logs, and review samples as evidence.
Record JSON · Open in map · Data retrieval guide
Catalog revision: 24028ffcfc2b295fa1b08ee6caa84b765f0731b321496bf4f548c49ad2177028. A connection does not establish full coverage.
Attributes
- domain
- AI Governance
- type
- preventive
- category
- technical
Details
- unified_id
- UC-AI-20
- title
- Prevent harmful, out-of-scope, hallucinated, and over-exposed AI outputs
- statement
- Filter and shape every AI output before release: block or transform content that matches the system's harmful-output taxonomy, keep responses within the declared scope and capabilities, detect agent-specific high-risk outputs and route them to defined responses by severity, ground factual claims in cited sources and verify them to limit hallucination, withhold system prompts, internal data, and other over-exposed information, and sanitize outputs consumed by downstream systems so they cannot carry executable or injected payloads. Measure filter effectiveness and retain configurations, block logs, and review samples as evidence.
- domain
- AI Governance
- control_type
- preventive
- control_category
- technical
- members
- framework
- aiuc-1
- control_id
- B009
- coverage
- full
- relationship
- superset_of
- framework
- aiuc-1
- control_id
- C003
- coverage
- full
- relationship
- superset_of
- framework
- aiuc-1
- control_id
- C004
- coverage
- full
- relationship
- superset_of
- framework
- aiuc-1
- control_id
- C005
- coverage
- full
- relationship
- superset_of
- framework
- aiuc-1
- control_id
- C006
- coverage
- full
- relationship
- superset_of
- framework
- aiuc-1
- control_id
- D001
- coverage
- full
- relationship
- superset_of
- guidance
- source
- nist-ai-tevv-athlon
- sourceTitle
- NIST AI 200-2: TEVV-Athlon Framework for Evaluating AI Systems
- propositionId
- NIST-TEVV-04
- propositionTitle
- Test for disclosure of confidential information
- sourcePages
- NIST AI 200-2 ipd Appendix B, Table 4, p. 24: Confidentiality attacks
Source
No record-specific source URL is provided.
Connections
- UC-AI-20 — Prevent harmful, out-of-scope, hallucinated, and over-exposed AI outputs maps_to C004 — Prevent out-of-scope outputs
- framework
- aiuc-1
- control_id
- C004
- coverage
- full
- relationship
- superset_of
- delta
- Not provided
- source_version
- July 15, 2026 release (quarterly update cadence)
- provenance
- mapper
- coworkcanvas-compliance-graph
- reviewDate
- 2026-09-07
- direction
- canonical_to_source
- defaultConfidence
- medium
- defaultStatus
- active
- note
- Each member is a documented relationship claim from the canonical unified control to a source control or guidance proposition. relationship: equal|superset_of (full) / intersects_with|subset_of (partial) / informs (guidance). confidence 'medium' = single-mapper, documented, not yet externally corroborated. source_version is the member framework's edition from the standard version register.
- UC-AI-20 — Prevent harmful, out-of-scope, hallucinated, and over-exposed AI outputs mitigates Harmful AI bias and discrimination against protected groups
- strength
- related
- rationale
- Harmful-output categories typically include discriminatory content, so output filtering catches some biased responses.
- UC-AI-20 — Prevent harmful, out-of-scope, hallucinated, and over-exposed AI outputs mitigates Inaccurate, unreliable or hallucinated AI outputs
- strength
- primary
- rationale
- Grounding, citation, and verification of factual claims directly reduce hallucinated and unreliable outputs.
- UC-AI-20 — Prevent harmful, out-of-scope, hallucinated, and over-exposed AI outputs mitigates Misuse of AI systems for offensive cyber operations or catastrophic harm
- strength
- related
- rationale
- Scope and harmful-content filters catch part of the misuse surface; the dedicated misuse refusal control is primary.
- AI Guardrail Configuration & Agent Permission Review operates UC-AI-20 — Prevent harmful, out-of-scope, hallucinated, and over-exposed AI outputs
- UC-AI-20 — Prevent harmful, out-of-scope, hallucinated, and over-exposed AI outputs maps_to D001 — Prevent hallucinated outputs
- framework
- aiuc-1
- control_id
- D001
- coverage
- full
- relationship
- superset_of
- delta
- Not provided
- source_version
- July 15, 2026 release (quarterly update cadence)
- provenance
- mapper
- coworkcanvas-compliance-graph
- reviewDate
- 2026-09-07
- direction
- canonical_to_source
- defaultConfidence
- medium
- defaultStatus
- active
- note
- Each member is a documented relationship claim from the canonical unified control to a source control or guidance proposition. relationship: equal|superset_of (full) / intersects_with|subset_of (partial) / informs (guidance). confidence 'medium' = single-mapper, documented, not yet externally corroborated. source_version is the member framework's edition from the standard version register.
- UC-AI-20 — Prevent harmful, out-of-scope, hallucinated, and over-exposed AI outputs maps_to B009 — Limit output over-exposure
- framework
- aiuc-1
- control_id
- B009
- coverage
- full
- relationship
- superset_of
- delta
- Not provided
- source_version
- July 15, 2026 release (quarterly update cadence)
- provenance
- mapper
- coworkcanvas-compliance-graph
- reviewDate
- 2026-09-07
- direction
- canonical_to_source
- defaultConfidence
- medium
- defaultStatus
- active
- note
- Each member is a documented relationship claim from the canonical unified control to a source control or guidance proposition. relationship: equal|superset_of (full) / intersects_with|subset_of (partial) / informs (guidance). confidence 'medium' = single-mapper, documented, not yet externally corroborated. source_version is the member framework's edition from the standard version register.
- UC-AI-20 — Prevent harmful, out-of-scope, hallucinated, and over-exposed AI outputs maps_to C003 — Prevent harmful outputs
- framework
- aiuc-1
- control_id
- C003
- coverage
- full
- relationship
- superset_of
- delta
- Not provided
- source_version
- July 15, 2026 release (quarterly update cadence)
- provenance
- mapper
- coworkcanvas-compliance-graph
- reviewDate
- 2026-09-07
- direction
- canonical_to_source
- defaultConfidence
- medium
- defaultStatus
- active
- note
- Each member is a documented relationship claim from the canonical unified control to a source control or guidance proposition. relationship: equal|superset_of (full) / intersects_with|subset_of (partial) / informs (guidance). confidence 'medium' = single-mapper, documented, not yet externally corroborated. source_version is the member framework's edition from the standard version register.
- UC-AI-20 — Prevent harmful, out-of-scope, hallucinated, and over-exposed AI outputs informed_by NIST-TEVV-04 — Test for disclosure of confidential information
- framework
- nist-ai-tevv-athlon
- control_id
- NIST-TEVV-04
- coverage
- guidance
- relationship
- informs
- delta
- Not provided
- source_version
- NIST AI 200-2 ipd (Initial Public Draft), August 2026
- provenance
- mapper
- coworkcanvas-compliance-graph
- reviewDate
- 2026-09-07
- direction
- canonical_to_source
- defaultConfidence
- medium
- defaultStatus
- active
- note
- Each member is a documented relationship claim from the canonical unified control to a source control or guidance proposition. relationship: equal|superset_of (full) / intersects_with|subset_of (partial) / informs (guidance). confidence 'medium' = single-mapper, documented, not yet externally corroborated. source_version is the member framework's edition from the standard version register.
- sourcePages
- NIST AI 200-2 ipd Appendix B, Table 4, p. 24: Confidentiality attacks
- UC-AI-20 — Prevent harmful, out-of-scope, hallucinated, and over-exposed AI outputs maps_to C006 — Prevent output vulnerabilities
- framework
- aiuc-1
- control_id
- C006
- coverage
- full
- relationship
- superset_of
- delta
- Not provided
- source_version
- July 15, 2026 release (quarterly update cadence)
- provenance
- mapper
- coworkcanvas-compliance-graph
- reviewDate
- 2026-09-07
- direction
- canonical_to_source
- defaultConfidence
- medium
- defaultStatus
- active
- note
- Each member is a documented relationship claim from the canonical unified control to a source control or guidance proposition. relationship: equal|superset_of (full) / intersects_with|subset_of (partial) / informs (guidance). confidence 'medium' = single-mapper, documented, not yet externally corroborated. source_version is the member framework's edition from the standard version register.
- Quarterly Third-Party AI Evaluation Cycle tests UC-AI-20 — Prevent harmful, out-of-scope, hallucinated, and over-exposed AI outputs
- UC-AI-20 — Prevent harmful, out-of-scope, hallucinated, and over-exposed AI outputs mitigates AI safety failures causing physical or psychological harm
- strength
- primary
- rationale
- Output filtering against the harmful-output taxonomy is the last technical barrier before a harmful response reaches a person.
- UC-AI-20 — Prevent harmful, out-of-scope, hallucinated, and over-exposed AI outputs maps_to C005 — Prevent agent-specific high risk outputs
- framework
- aiuc-1
- control_id
- C005
- coverage
- full
- relationship
- superset_of
- delta
- Not provided
- source_version
- July 15, 2026 release (quarterly update cadence)
- provenance
- mapper
- coworkcanvas-compliance-graph
- reviewDate
- 2026-09-07
- direction
- canonical_to_source
- defaultConfidence
- medium
- defaultStatus
- active
- note
- Each member is a documented relationship claim from the canonical unified control to a source control or guidance proposition. relationship: equal|superset_of (full) / intersects_with|subset_of (partial) / informs (guidance). confidence 'medium' = single-mapper, documented, not yet externally corroborated. source_version is the member framework's edition from the standard version register.
- UC-AI-20 — Prevent harmful, out-of-scope, hallucinated, and over-exposed AI outputs mitigates AI privacy leakage and re-identification
- strength
- related
- rationale
- Withholding internal data and over-exposed content from outputs reduces personal-data leakage at the response boundary.