# Workforce AI Security Requirements (Harmonic Security)

> 69 vendor-evaluation requirements for platforms that secure workforce use of AI, each with acceptance criteria, a verification test and explained mappings to 6 frameworks. Updated October 2026.

Mappings indicate relevant contributions, not compliance. Every page is also available as markdown by appending .md (and JSON with .json). Full text: /llms-full.txt.

## Data
- [All requirements (markdown)](/llms-full.txt): every requirement with mappings
- [Requirements JSON](/api/requirements.json): filter with ?area=, ?framework=, ?q=
- [Areas and frameworks JSON](/api/frameworks.json): slugs and counts
- [Requirements CSV](/api/requirements.csv): spreadsheet with a column per framework

## Capability areas
- [AI Security Posture Management](/SPM.md): 13 requirements. Discover every AI tool, account, model and agent in use, and keep posture continuously assessed.
- [Runtime Guardrails](/GRD.md): 25 requirements. Inspect prompts, files and responses in real time; coach, redact or block before data leaves.
- [AI Detection & Response](/AIDR.md): 10 requirements. Detect risky AI behavior, investigate incidents and respond with full context.
- [Agent & MCP Governance](/AGT.md): 9 requirements. Govern agents, MCP servers and tool calls with identity, scope and runtime control.
- [AI Usage, Spend & Value](/VAL.md): 9 requirements. Measure adoption, spend and business value so AI investment decisions are evidence-led.
- [Audit & Evidence](/AUD.md): 7 requirements. Produce tamper-evident logs and auditor-ready evidence across every control.
- [Platform, Privacy & Assurance](/PLT.md): 15 requirements. Deployment, data residency, privacy-by-design and vendor assurance.

## Frameworks
- [ISO/IEC 27001:2022](/ISO27001.md): 65 mapped requirements. Information security management system, including Annex A controls (Amd 1:2024 noted).
- [ISO/IEC 42001:2023](/ISO42001.md): 54 mapped requirements. AI management system standard, including Annex A AI controls.
- [NIST Cybersecurity Framework 2.0](/NIST-CSF.md): 64 mapped requirements. Govern, Identify, Protect, Detect, Respond and Recover outcomes.
- [NIST AI Risk Management Framework](/NIST-AI-RMF.md): 54 mapped requirements. Govern, Map, Measure and Manage functions for trustworthy AI.
- [SOC 2 Trust Services Criteria](/SOC2.md): 65 mapped requirements. AICPA 2017 Trust Services Criteria with revised points of focus (2022).
- [OWASP Top 10 for Agentic Applications](/OWASP-ASI.md): 10 mapped requirements. Agentic security risks ASI01–ASI10, from goal hijack to rogue agents.

## Requirements
- [SPM-001: Classify previously unseen AI applications on first encounter](/r/SPM-001.md)
- [SPM-002: Detect AI use at the browser point of use](/r/SPM-002.md)
- [SPM-003: Detect endpoint AI activity beyond the browser](/r/SPM-003.md)
- [SPM-004: Detect direct model-provider API calls from employee devices](/r/SPM-004.md)
- [SPM-005: Discover local model runtimes and offline AI installations](/r/SPM-005.md)
- [SPM-006: Discover AI embedded in sanctioned SaaS](/r/SPM-006.md)
- [SPM-007: Provide visibility for unmanaged endpoints](/r/SPM-007.md)
- [SPM-008: Distinguish personal, free and guest accounts from sanctioned enterprise tenants](/r/SPM-008.md)
- [SPM-009: Risk-score AI tools, connectors and agent workflows](/r/SPM-009.md)
- [SPM-010: Define default treatment for newly discovered AI services](/r/SPM-010.md)
- [SPM-011: Support the AI tool approval and review lifecycle](/r/SPM-011.md)
- [GRD-001: Inspect prompt content in real time before submission](/r/GRD-001.md)
- [GRD-002: Inspect file uploads before submission](/r/GRD-002.md)
- [GRD-003: Classify sensitive information semantically in ordinary business language](/r/GRD-003.md)
- [GRD-004: Detect and coach in the languages your workforce uses](/r/GRD-004.md)
- [GRD-005: Detect general, customer, employee and bulk PII](/r/GRD-005.md)
- [GRD-006: Detect sensitive health and workforce information](/r/GRD-006.md)
- [GRD-007: Detect confidential commercial information](/r/GRD-007.md)
- [GRD-008: Detect secrets, technical configuration and security-operations detail](/r/GRD-008.md)
- [GRD-009: Detect proprietary source code and engineering content](/r/GRD-009.md)
- [GRD-010: Support granular custom detections](/r/GRD-010.md)
- [GRD-011: Maintain acceptable detection quality](/r/GRD-011.md)
- [GRD-012: Apply policy by user, group, role, department, tool, account plan and risk context](/r/GRD-012.md)
- [GRD-013: Support monitor, warn, block, and restrict actions](/r/GRD-013.md)
- [GRD-014: Prevent sensitive data reaching unauthorized tools or account tiers](/r/GRD-014.md)
- [GRD-015: Coach employees in the moment](/r/GRD-015.md)
- [GRD-016: Explain each decision to the employee and the analyst](/r/GRD-016.md)
- [GRD-017: Provide sensible policy defaults and no-code authoring](/r/GRD-017.md)
- [GRD-018: Turn acceptable-use policy into enforceable scenarios](/r/GRD-018.md)
- [GRD-019: Roll out policy in stages, starting in monitor-only mode](/r/GRD-019.md)
- [GRD-020: Scope exceptions to a single detection for a defined group](/r/GRD-020.md)
- [GRD-021: Keep typical added inline latency under 300 ms](/r/GRD-021.md)
- [IDR-001: Generate actionable alerts](/r/IDR-001.md)
- [IDR-002: Detect direct and indirect prompt injection attempts.](/r/IDR-002.md)
- [IDR-003: Identify use cases leading to elevated risk behavior.](/r/IDR-003.md)
- [IDR-004: Integrate with SIEM, SOAR, ticketing, and identity systems](/r/IDR-004.md)
- [AGT-001: Govern tool calls at read, write and delete scope](/r/AGT-001.md)
- [AGT-002: Flag destructive, over-scoped or anomalous agent actions](/r/AGT-002.md)
- [AGT-003: Inspect sensitive data retrieved through connectors mid-workflow](/r/AGT-003.md)
- [AGT-004: Attribute every connector and tool call alert to a person](/r/AGT-004.md)
- [AGT-005: Provide targeted containment for a single connector or tool.](/r/AGT-005.md)
- [AGT-006: Discover the MCP servers and connectors employees attach to their AI clients](/r/AGT-006.md)
- [AGT-007: Map the actions available via each connector.](/r/AGT-007.md)
- [AGT-008: Support local and remote MCP gateway placement for employee AI clients](/r/AGT-008.md)
- [AGT-009: Detect indirect prompt injection in content agents read](/r/AGT-009.md)
- [VAL-001: Measure AI usage and adoption by tool, team, function and workflow](/r/VAL-001.md)
- [VAL-002: Classify AI activity into business-relevant use cases](/r/VAL-002.md)
- [VAL-003: Measure spend, waste and portfolio decisions](/r/VAL-003.md)
- [VAL-004: Measure productivity and business-value signals](/r/VAL-004.md)
- [VAL-005: Generate targeted training and enablement recommendations](/r/VAL-005.md)
- [VAL-006: Measure employee self-remediation](/r/VAL-006.md)
- [VAL-007: Provide role-based dashboards and reporting, up to board level](/r/VAL-007.md)
- [AUD-001: Keep a configurable record of each conversation](/r/AUD-001.md)
- [AUD-002: Export audit and forensic evidence](/r/AUD-002.md)
- [AUD-003: Restrict and audit access to raw prompt content](/r/AUD-003.md)
- [PLT-001: Minimize endpoint and network architecture risk](/r/PLT-001.md)
- [PLT-002: Disclose where decisions are enforced and how the control fails](/r/PLT-002.md)
- [PLT-003: Deploy silently through endpoint management with release controls](/r/PLT-003.md)
- [PLT-004: Support Linux endpoints](/r/PLT-004.md)
- [PLT-005: Support network-based AI traffic detection where the architecture requires it](/r/PLT-005.md)
- [PLT-006: Scale enforcement and telemetry to the full population](/r/PLT-006.md)
- [PLT-007: Resist common circumvention paths](/r/PLT-007.md)
- [PLT-008: Separate administrative roles, and authenticate them through enterprise SSO](/r/PLT-008.md)
- [PLT-009: Control telemetry residency, retention, encryption, deletion and masking](/r/PLT-009.md)
- [PLT-010: Offer required hosting regions](/r/PLT-010.md)
- [PLT-011: Inspect uploaded content without permanent file retention](/r/PLT-011.md)
- [PLT-012: Support pseudonymized and aggregate-only views where required](/r/PLT-012.md)
- [PLT-013: Limit monitoring scope by design](/r/PLT-013.md)
- [PLT-014: Maintain security assurance](/r/PLT-014.md)

## Optional
- [Methodology and sources](/methodology)
- [Agent conventions](/agents)