Security & Compliance Pack
Comprehensive security documentation for procurement, compliance, and information security teams reviewing HybridLLM.
1. Scope & Deployment Modes
HybridLLM is available in two deployment modes: (1) SaaS: Swiss-hosted managed service operating exclusively on Swiss infrastructure within Tier III+ certified data centers. All compute, storage, and networking remain within Swiss jurisdiction. (2) Enterprise: Customer-deployed installation in the customer's Swiss infrastructure (Swiss cloud / Swiss data center). Specific deployment parameters are documented during procurement.
2. Security Controls Overview
Encryption: AES-256 for data at rest; TLS 1.3 for data in transit. Encryption keys are managed using hardware security modules (HSMs) in SaaS deployments. Access Control: Multi-factor authentication (MFA) and Single Sign-On (SSO) via SAML 2.0 are supported for Enterprise plans. Monitoring: 24/7 automated threat detection with defined incident response procedures. Notification SLA for security breaches: <1 hour for critical incidents, <24 hours for non-critical events.
3. Governance: Roles & Workspace Policies
Role-based access control (RBAC) enforces least-privilege access. Roles include: Owner with full system control; Admin with workspace management and user provisioning; Member with document upload and query access. Workspace policies control optional features such as web retrieval and anonymization. Admins configure these policies per workspace. Audit logs capture all policy changes.
4. Optional Web Retrieval & Anonymization
Web Retrieval: Optional feature controlled by workspace policy. When enabled, HybridLLM retrieves real-time information from the web to augment responses. Web retrieval is provided by a SERP API provider (provider may vary; documented in DPA annex). Anonymization: Optional internal LLM-based feature that redacts personally identifiable information (PII) from prompts before processing. Anonymization occurs within HybridLLM infrastructure and does not involve external services. Both features are disabled by default and require explicit workspace-level enablement.
5. Enterprise Optional Public AI Augmentation
Enterprise deployments support optional public AI augmentation as a custom enablement feature. This allows routing of specific queries to external AI providers when explicitly enabled by the customer. Customer controls: prompt preview before submission; anonymization toggle (optional redaction of PII before external transmission); optional document anonymization. This feature is disabled by default and requires custom configuration. Customer retains full control over enablement, provider selection, and data transmission decisions.
6. Auditability
Immutable audit logs capture all user actions including queries, document uploads, exports, policy changes, and administrative operations. Logs include timestamps, user IDs, IP addresses, session metadata, and workspace context. Retention policies are configurable per workspace (90 days to unlimited). Audit logs can be exported to SIEM tools (Splunk, Datadog, Azure Sentinel) for centralized compliance monitoring. Deletion requests are logged and verifiable.
7. Subprocessors
SaaS: Infrastructure is provided exclusively by Switzerland-based subcontractors (Hidora for hosting). No customer data, prompts, or embeddings are shared with third-party AI services or processors outside Switzerland. Web retrieval (when enabled) is provided by a SERP API provider (category: search engine results provider; specific provider may vary and is documented in DPA annex). Enterprise: Subprocessor list depends on customer deployment configuration and infrastructure choices. Enterprise subprocessors are documented in the Order Form / DPA annex. A complete subprocessor list is available in the Data Processing Agreement (DPA) and updated quarterly.
8. Customer Responsibilities
Customers are responsible for: (1) Configuring workspace policies for optional features (web retrieval, anonymization, public AI augmentation); (2) Governing usage of optional outbound features and making informed decisions about data transmission; (3) Evaluating anonymization effectiveness for their specific use cases; (4) Ensuring compliance with applicable data protection regulations when enabling optional external features. HybridLLM provides technical controls and audit logs to support customer governance decisions.
9. Evidence Available Under NDA
HybridLLM maintains comprehensive security documentation available to qualified customers under Non-Disclosure Agreement (NDA). Available evidence includes: independent security testing summaries; remediation evidence for identified vulnerabilities; infrastructure architecture diagrams; detailed audit log specifications; external audit engagement roadmap (timing and scope subject to business planning). Customers may request evidence packages by contacting [email protected]. Evidence requests are typically fulfilled within 5 business days.
Request Documentation Package
Full technical security documentation, independent testing summaries, and detailed architecture diagrams are available under NDA for compliance and procurement teams.