01
Current posture
This page separates controls implemented in ELOHM’s service contracts and preview logic from controls that require a production deployment. It is a candid description of posture, not a certification claim. The public website and product preview do not provide a live customer tenant or production runtime.
- Enforcement-backed in the service layer: classification, permissions, minimum-necessary context, provider policy, and approval decisions
- Requires production infrastructure: live identity, hardened storage, network controls, deployment monitoring, backups, and operating procedures
02
Tenant isolation
ELOHM carries organization, company, and workspace identifiers with scoped records. A cross-company view requires an explicitly permitted scope and an authority reference. In production, those decisions must be backed by the customer’s identity provider, server-side repositories, isolated storage, and deployment controls. The preview demonstrates the boundary model locally; it is not evidence of a live tenant boundary.
- Organization → company → workspace scope
- Explicit cross-scope authorization
- Browser-local preview identity is not production authentication
03
Permissions and data scope
Access decisions are intended to consider the active user, organization, workspace, role, permission, source, and action. ELOHM should show the context that is in scope rather than silently treating every connected source as available. Writes and external effects remain subject to narrower authority and approval boundaries.
- Scope is part of the request, not hidden background state
- Read, analyze, recommend, and act are distinct boundaries
- Customers remain responsible for role and source configuration
04
Authority and approval boundaries
ELOHM is designed to support human judgment. Recommendations are not automatically executed actions. Guardians, permissions, policy settings, and approval states are intended to stop a request when the authority or required approval is missing. Production deployments must connect those decisions to an accountable identity and an operational change path.
- External effects require stronger checks than analysis
- Human approval stays visible for consequential actions
- A blocked request should explain the boundary that stopped it
05
Document sensitivity
Document and conversation context is classified as internal, confidential, sensitive, or restricted based on the information detected and the source context available to ELOHM. Restricted examples include credentials, private keys, identity numbers, and health records. Classification is a control input, not a guarantee that an automated classifier will understand every document; customers should use explicit source policy and review.
- Document-level and request-level sensitivity signals
- Minimum necessary excerpts rather than unrestricted transmission
- Customer review remains necessary for ambiguous material
06
Audit chain and traceability
ELOHM’s service model carries provenance, source, organization, company, workspace, role, freshness, workflow, and audit information so that an answer or action can be inspected. A production audit chain also needs durable server-side storage, time synchronization, access controls, retention, export, and monitoring. Those infrastructure pieces are deployment responsibilities until connected.
- Source and provenance should travel with organizational knowledge
- Actions and policy decisions should be inspectable
- Preview activity is illustrative and browser-local
07
Secrets and external calls
API keys and provider credentials belong on the server side. They must never be embedded in frontend source, exposed through browser storage, or passed to a page for convenience. Before an external API call, ELOHM’s AI Collaboration Gateway is intended to evaluate data classification, permissions, provider policy, and minimum necessary context. External collaborators are optional and can be blocked by organization policy.
- Server-side secrets handling
- Classification and permission checks before external calls
- Sovereign Mode: ELOHM-owned models and infrastructure, with no data leaving the ELOHM environment
08
Incident response
A production incident process should include detection, triage, containment, evidence preservation, customer communication, remediation, and a post-incident review. The exact response time, notification threshold, and contact path belong in the applicable customer agreement and deployment runbook. This preview does not claim a staffed 24/7 response operation.
- Report suspected security issues promptly
- Do not include secrets or exploit details in a public form
- Use the contact form and select Security for responsible disclosure
09
Responsible disclosure
To report a suspected security issue, use the Contact page and choose Security in the interest field. Include a concise description, affected page or behavior, reproduction steps that do not access another person’s data, and a safe way to follow up. Do not include API keys, passwords, personal data, or customer content in the form. ELOHM by Durandisse Industries will review the report and coordinate a response when production contact operations are available.
- Security disclosure: /contact
- Please provide safe reproduction details
- No direct security mailbox is published in this preview