Resilient Cloud Architecture
Production workloads run on hardened cloud-native infrastructure with multi-zone deployment, encrypted storage, and tightly scoped administrative access.
Your compliance evidence deserves the same rigor you put into it. Here is how we protect it.
Last Reviewed: May 2026
Documented controls across CC1 to CC9. SOC 2 Type I scope is Security, one of five AICPA categories.
Internal DAST engagement.
ProfytAI's production environment uses cloud-native infrastructure with network segmentation, environment segregation, encrypted storage, managed key lifecycle, and recoverable backup. Each layer limits blast radius under failure or compromise.
Production workloads run on hardened cloud-native infrastructure with multi-zone deployment, encrypted storage, and tightly scoped administrative access.
Public ingress is isolated from protected application and data tiers. Sensitive endpoints sit behind authentication, authorization, and continuous traffic inspection.
Production and non-production environments are logically separated to contain cross-environment risk and support controlled evidence handling.
Backups are encrypted with managed key controls, retained against documented policy, replicated cross-region, with documented recovery procedures aligned to business continuity requirements.
Encryption and Key Management
ProfytAI applies industry-standard cryptographic primitives across data at rest, data in transit, key lifecycle management, and transport-layer hardening.
Customer data, backups, and platform storage are encrypted at rest using AES-256 with managed key lifecycle controls.
Client and service communications use TLS 1.2 or higher, with TLS 1.3 supported. HTTP requests are 301-redirected to HTTPS at the edge, and HSTS is preload-eligible and submitted to the Chrome preload list.
SSLv2, SSLv3, TLS 1.0, and TLS 1.1 are not offered. The public web tier holds an A+ grade from Qualys SSL Labs across all CloudFront edge nodes (assessed May 2026).
ProfytAI's control environment is organized against the AICPA Trust Services Criteria for Security. Twenty-two documented controls are mapped to the Common Criteria CC1 through CC9 and organized for future cross-walk to ISO/IEC 27001:2022 Annex A.
Documented controls are organized against the AICPA Trust Services Criteria for Security (TSP Section 100, Common Criteria CC1 through CC9), covering governance, access, change management, system operations, monitoring, and risk mitigation.
Customer data safeguards are documented around ownership, confidentiality, least-privilege use, encrypted handling, and controlled evidence release.
Monitoring, incident handling, continuity planning, and vulnerability workflows support secure operation of the customer-facing platform.
Workforce access follows the principle of least privilege under enforced multi-factor authentication, group-based authorization, and centralized audit logging. Product delivery runs through peer review, automated pre-merge quality gates, and continuous secrets scanning.
Every change to the production branch requires peer review and approval before merge, with branch protection enforced at the source-control layer.
Build, static type checking, lint, and the regression test suite run automatically before any change can reach production. Failed gates block the merge.
Branch protection on production, release, and hotfix branches blocks force pushes and requires pull-request approval. Conventional-commit format and release notes preserve a structured change history for change traceability and incident reconstruction during audit.
Secrets are scanned pre-commit and continuously in source. Production deploys are operated by named, authorized engineering personnel only.
ProfytAI separates the regulator's text from interpretation. The regulator's wording and page always come from the source document, never from AI. AI fills only the judgment fields, sees only public regulatory text, and is always anchored to the source paragraph.
Gap Analysis is a separate path. After PII redaction, your uploaded policies are scored by ProfytAI's internal cloud agents inside our controlled environment only and never leave our network.
The regulator's wording and page always come from the source document, never from AI, with machine-readable document-path identifiers that regulators can verify independently.
AI fills the judgment fields, such as who must act and under which conditions, against public regulatory text only. Every downstream artifact carries a permanent source anchor back to the verbatim source paragraph, and model output is flagged for human review.
Material compliance decisions are reviewed and approved by the customer's compliance team. ProfytAI does not auto-publish, auto-attest, or auto-submit to any regulator.
Prompts and documents used for inference stay inside our controlled cloud environment. They are not sent to external model vendors or public AI APIs. This is an architectural property, not a policy preference.
Customer prompts and outputs are contractually barred from model training. Model versions are pinned to specific releases, and upgrades follow documented regression testing.
ProfytAI operates continuous security monitoring across the platform. Threat telemetry, business continuity awareness, and vulnerability management procedures support secure operation of the customer-facing platform.
Cloud-native anomaly-detection signals trigger alerts that feed into our incident response process.
Operational continuity planning, multi-zone resilience, and cross-region backup support service resilience and recovery coordination.
Security findings are triaged, prioritized, and tracked to remediation through a documented risk-based workflow.
Security overview, SOC 2 readiness materials, internal assessment summaries, and data-protection documents are released to qualified reviewers under appropriate confidentiality controls. Security concerns are handled through coordinated vulnerability disclosure per RFC 9116.
Assurance Evidence
Summary of governance, encryption, access management, monitoring, and operational resilience controls.
Readiness evidence mapped to the AICPA Trust Services Criteria for Security, including documented control coverage.
Internal automated DAST findings from the public web surface, provided to qualified reviewers under NDA without exposing non-public test material.
Customer-data handling, encryption, retention, ownership, confidentiality, and processing commitments.
Coordinated Vulnerability Disclosure
Coordinated through RFC 9116 and the ProfytAI security channel. Out-of-scope activity includes denial-of-service testing, phishing, credential attacks, and any attempt to access customer or third-party data.
Customers and partners route security concerns through their established ProfytAI support or security-review channel. Researchers without one use the contact form linked from the published security contact discovery file, which is also what automated scanners read.
Reports include enough detail to reproduce and triage the concern without exposing confidential data in public forums or non-secure channels.
ProfytAI will not pursue legal action against security researchers acting in good faith under this policy. Stay within scope, avoid privacy violations and service disruption, and give us a reasonable window to remediate before public disclosure.
We acknowledge eligible reports within 5 business days and provide status updates at least every 14 days until the issue is resolved.