Security enforced at every layer.
Every decision and audit record is tamper-evident and independently re-verifiable, API access is least-privilege and scoped, outbound events are signed and SSRF-hardened, and every change is gated on a passing safety-guard suite. Security is enforced in code and CI — and the platform has no capital-movement path to begin with.
AI analyzes. Deterministic systems validate. Humans authorize. Capital remains governed.
The controls that run today.
Concrete, verifiable controls in place across records, access, delivery, change management, and the software supply chain — each stated as what it does, not what it aspires to.
Decision and audit records are append-only with a per-record provenance hash, enforced at the database layer. A recompute-on-read verify endpoint lets a reviewer re-check integrity independently — evidence you can audit, not just trust.
Restricted, workspace-scoped API keys confine a key to a route allow-set. Keys are stored as one-way hashes, shown once at creation, and compared in constant time.
Outbound events are HMAC-signed and delivered through an SSRF-hardened gate: private, loopback, and metadata-address ranges are refused and DNS-rebinding is re-checked at send time. Delivery stays off until explicitly enabled.
The protected main branch requires review and passing continuous-integration checks — including an automated safety-guard suite — before any change can merge. Security posture is enforced in code and CI, not documented after the fact.
Dependency updates flow through reviewed pull requests, and an architecture- and safety-guard test suite runs on every merge to main. A static-analysis (SAST) pipeline (CodeQL) is configured and pending activation — we do not represent it as running until it is.
Public traffic is TLS-encrypted with strict transport-security and content-security policies, and a published vulnerability-disclosure policy (security.txt, RFC 9116) is served on every host.
The strongest control is a capability you never expose.
SigmaX prepares evidence and decision records — it has no path to turn them into capital movement. Removing that capability removes an entire class of breach: there is nothing to compromise, misconfigure, or exfiltrate into a transfer.
No broker, custody, transfer, allocation, or customer-capital routing surface exists in the product. The class of breach that moves money is designed out, not merely disabled.
AI, Atlas, the website, and the client portal produce evidence, proof gaps, approval state, and route state. They cannot authorize a capital-linked action — a human is always the authorization layer.
Provider credentials are operator-owned and never exposed on public surfaces, which reference env-var names and readiness — never secret values.
Sample, demo, paper, and real evidence stay labeled at every layer. Sample data cannot become real customer evidence by presentation alone.
Impact Ledger and AuditWriter preserve who reviewed what, which evidence existed, and what route state was produced — a reconstructable trail for any diligence request.
Any future live-execution or customer-capital route is gated behind separate legal review and explicit operator approval before it can exist.
Status stated exactly — never implied.
In-product evidence lineage is available to review today. Formal third-party attestation is a separate line — we state its status precisely and never imply coverage we do not have.
Decision and evidence lineage can be assembled for any review as part of the Capital Action OS workflow. These records inform decisions; they do not grant execution authority.
Demo outputs, simulations, and sample packets stay visibly labeled and cannot become real customer evidence by presentation alone.
We do not represent SOC 2 or an external penetration test as finished. When an artifact exists, this page will state its exact status or link the report — nothing sooner.
A formal public subprocessor register is not yet published. Current subprocessors can be shared under a diligence review rather than implied as a complete list.
Security posture by access path.
Builder Sandbox
Learns the ActionIntent lifecycle without live authority, live customer capital, or external execution.
Authority Team and Institutional OS
Supports governed review records, required-approver state, mandate checks, route status, and audit output.
ActionIntent API
Returns decision-check results and references through least-privilege scoped keys — without becoming a broker, custodian, transfer system, or order router.
Private Infrastructure and Pilot
Scopes private controls, evidence posture, and implementation requirements before any production deployment.
SigmaX provides capital decision intelligence, scenario analysis, and authority workflow software. SigmaX does not provide investment advice. SigmaX does not place trades, move funds, route customer capital, or manage client capital. Shadow/simulated/paper/backtested results are not live trading results. No result is a guarantee of future performance. Human approval is required before any capital-linked action.
Where SigmaX uses AI, it is governed as assistive software: truthful capability claims, human oversight, traceable records, disclosure of AI assistance, and no capital authority. SigmaX designs AI controls with reference to applicable securities obligations, FINRA existing-rule guidance, SEC AI-claims discipline, NIST AI RMF-style risk management, and EU AI Act transparency readiness. This is compliance-readiness design, not a claim that every deployment or jurisdiction-specific use has completed legal review.