โฌก SOC 2-Aligned

Security & Trust Controls

WorkDecisions AI operates under a security framework aligned with AICPA SOC 2 Trust Service Criteria (TSC). Our controls address the five TSC categories: Security, Availability, Processing Integrity, Confidentiality, and Privacy. This document describes those controls, their operational status, and evidence of implementation.

Note on SOC 2 status: WorkDecisions AI operates SOC 2-aligned controls. We have not yet completed a formal SOC 2 Type II audit by an accredited CPA firm. Formal SOC 2 Type II audit is targeted as part of our enterprise client roadmap. We do not misrepresent our audit status. Prospects requiring a completed SOC 2 report may contact us to discuss our controls evidence directly.

๐Ÿ›๏ธ Trust Service Criteria Coverage

SOC 2 evaluates controls across five Trust Service Criteria. The table below shows our coverage status against each criterion.

CC1โ€“CC9 ยท Security
Common Criteria (CC)
Logical and physical access controls, change management, risk assessment, monitoring. Our primary focus. All CC series controls implemented.
A-series ยท Availability
Availability
System availability for operation and use as committed. Supabase SLA 99.9%. Netlify SLA 99.99%. Monitoring and incident escalation active.
PI-series ยท Processing Integrity
Processing Integrity
System processing is complete, valid, accurate, and timely. All AI decision outputs are logged with actor ID, timestamp, and input hash for audit traceability.
C-series ยท Confidentiality
Confidentiality
Information designated as confidential is protected. Multi-tenant RLS enforced at database layer. Client data never co-mingled. AES-256 at rest.

๐Ÿ” CC1โ€“CC2: Control Environment & Communication

LiveDocumented security policies reviewed at start of every operational session. Agents operating within the swarm are bound by documented protocols.
LiveAll external-facing communications reviewed and approved by CEO (Glen Allison) prior to send. No unsanctioned outreach permitted.
LiveSecurity responsibilities defined and assigned per agent and edge function role. Principle of least privilege enforced.
LiveHard Rules (HR11โ€“HR22) constitute the formal security policy framework, reviewed and updated after every material infrastructure change.
PlannedAnnual formal policy review with sign-off โ€” targeted Q4 2026 aligned with ISO audit readiness.

๐Ÿ”‘ CC6: Logical Access Controls

LiveEmail + Phone OTP two-factor authentication on all client portal access. No password-only routes exist.
LiveJWT tokens in HttpOnly cookies โ€” not exposed to JavaScript. XSS cannot steal authentication tokens.
LiveRow-Level Security (RLS) enforced at PostgreSQL layer โ€” every query scoped to tenant. Cross-client data access is architecturally blocked, not just application-enforced.
LiveAll secrets stored in Frankfurt edge function environment only. No secrets in source code, .env files, application variables, or third-party vaults. Agents receive key names, never values.
LiveAccess revocation: edge function access can be removed within 5 minutes via Frankfurt console. Database row-level access revocable without full system downtime.
PlannedQuarterly access review and privilege attestation โ€” Q3 2026.

๐Ÿ“Š CC7: System Operations & Monitoring

All significant system events are written to the Frankfurt audit_log table. Monitoring is continuous.

ControlDescriptionStatus
Event Logging Every authentication event, data access, compliance action, and AI output logged with timestamp, actor ID, IP address, and outcome. Live
Log Integrity Audit logs are append-only. No application-layer delete permissions exist on audit_log. Retention: 7 years for compliance events, 90 days for operational logs. Live
Anomaly Detection Automated monitoring on Frankfurt edge functions. Anomalous patterns (unusual query volume, failed auth bursts, unexpected data access) trigger SMS alert to CEO within 15 minutes. Live
Vulnerability Management Infrastructure dependencies reviewed quarterly. No known-vulnerable packages in production. Supabase, Netlify, and RunPod managed infrastructure with their own CVE programs. Live
Penetration Testing Independent penetration test โ€” targeted H2 2026. Planned

๐Ÿ”„ CC8: Change Management

LiveAll material infrastructure changes documented in CLAUDE.md, System Architecture, and Frankfurt audit_log before and after deployment.
LiveProduction deployments (workdecisionsai.com, workdecisions.live) require explicit CEO approval in writing before execute. No autonomous production deploys permitted.
LiveAll prospect email sends require explicit CEO approval. No autonomous outreach permitted.
LiveSandbox environment (wdi-portal-sandbox.netlify.app) maintained for all pre-approval testing. Production is only updated after sandbox is reviewed and approved.
LiveDeprecated workflows documented and never reactivated. Production workflow list maintained and verified.

โš ๏ธ CC9: Risk Mitigation & Vendor Management

Third-party vendors and sub-processors are assessed before integration. Risk criteria include: data residency, certification status, breach history, DPA availability, and exit rights.

VendorRisk AssessmentMitigation
Supabase ISO 27001 certified. Frankfurt EU data residency. SOC 2 Type II certified. Primary infrastructure. GDPR-sovereign.
RunPod Switzerland jurisdiction โ€” GDPR Article 45 adequacy decision. AI compute only. Isolated containers per workload. No persistent client PII.
HeyGen US-based. Processes voice/video data. No ISO 27001. SCCs in DPA. Disclosed to clients before contract. Optional alternative available.
ElevenLabs US-based. Processes voice data. No ISO 27001. SCCs in DPA. Disclosed to clients before contract. Optional alternative available.
Netlify Global CDN. No client PII on CDN nodes. Static assets only. No personal data served via Netlify.
Resend US-based email delivery. Processes email addresses and content. SCCs in DPA. Transactional email only โ€” no marketing or bulk sends without consent.

Full sub-processor list with DPA references: GDPR & Data Protection page.

โšก Availability Controls

๐Ÿ”’ Confidentiality Controls

๐Ÿ›ก๏ธ Privacy Controls

Last updated: May 2026. Controls reviewed quarterly and following any material infrastructure change. For audit evidence requests or security questionnaires: glen@workdecisionsai.com