Skip to content

docs: align trust claims with code (SOC 2, raw data access, TrustGuard retention) - #526

Merged
vgmartinez merged 1 commit into
developfrom
fix/docs-trust-claims-alignment
Nov 16, 2026
Merged

vgmartinez merged 1 commit into
developfrom
fix/docs-trust-claims-alignment

Conversation

@telmolivella2

Copy link
Copy Markdown
Contributor

Aligns security, privacy and retention claims in the docs with what the code and our certifications actually support.

Changes

SOC 2 — neuraltrust/security/overview.mdx, neuraltrust/data-privacy/overview.mdx

  • We hold ISO 27001 but not SOC 2. SOC 2 Type II is now listed as audit in progress, no report issued yet instead of a current certification with annual audits.

"NeuralTrust cannot access your raw data" — same two pages

  • Removed the "zero-knowledge architecture" claims. On SaaS, raw payloads and the TrustGuard request bodies behind findings are stored by NeuralTrust (documented in trustguard/data-handling.mdx, "What is stored"). The pages now say data location depends on deployment, and that Hybrid is the option that keeps raw payloads out of NeuralTrust.
  • Rewrote the "Data Sovereignty Model" paragraph, which claimed all processing happens in the customer's infrastructure and contradicted its own SaaS/Hybrid table.

TrustGuard retention — trustguard/data-handling.mdx, platform/compliance/eu-ai-act.mdx, platform/compliance/iso-42001.mdx

  • "365 days by default" was only true for Enterprise. Per-plan windows come from app (app/constants/entitlements.ts): Free 7, Standard 30, Enterprise 365 days, stamped on each record by TrustGuard (internal/domain/guard/entitlements.go).
  • Documented the fallback when a guard has no retention window: findings 90 days, request bodies 30 days (DataCore TTL views, core-infrastructure/clickstack-otel-collector/CONTRACT.md → Retention). On prod 2026-10-07 most trustguard_events rows were unstamped and on the 90-day default.
  • ISO 42001 page no longer claims every record carries a stamped expiry.

Out of scope (not in this repo)

  • Privacy policy (neuraltrust.ai, Storyblok): says nothing about the ml__prod_data_analysis pipeline, which exports TrustGuard findings from SaaS ClickHouse and relabels them through the OpenAI batch API. Also lists the office as Llacuna 162 in the policy body but Llacuna 161 in the footer. Needs legal.

To confirm before merge

  • The SOC 2 wording ("audit in progress").
  • Whether we want to name the Standard plan publicly.

🤖 Generated with Claude Code

…d retention)

- SOC 2 Type II: audit in progress, no report yet; ISO 27001 stays certified.
- Drop "zero-knowledge / NeuralTrust cannot access your raw data": on SaaS,
  raw payloads and TrustGuard request bodies are stored by NeuralTrust.
- TrustGuard retention: per-plan windows (Free 7, Standard 30, Enterprise 365
  days) and the unstamped defaults (findings 90 days, request bodies 30 days).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@github-actions

Copy link
Copy Markdown

Trivy code scan — HIGH/CRITICAL

No HIGH or CRITICAL vulnerabilities found. (run)

@telmolivella2
telmolivella2 marked this pull request as ready for review October 30, 2026 08:42
@vgmartinez
vgmartinez merged commit 0fe0397 into develop Nov 16, 2026
9 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants