Verisa AI
Your data. Your infrastructure. Your AI.
Verisa AI runs entirely inside your perimeter — air-gapped if you need it. It combines retrieval, ASR, and graph-native reasoning so every answer traces back to a specific source.
Sovereign by configuration
Cloud, on-premise, government network, or air-gapped — same codebase, different settings flag. Phone-home blocking and outbound allowlisting are enforced at source-code and network level.
Graph-native legal reasoning
Conventional retrieval breaks legal cross-references and exception hierarchies. Verisa's graph reasoning preserves them — lex specialis, lex posterior, amendment chains, precedent links — and the graph path is the explanation.
Hybrid retrieval stack
Vector search and a knowledge graph for everyday content; tree-based hierarchical retrieval and graph-native reasoning for compliance-grade legal and regulatory work. One platform, four retrieval modes, mixable per agent.
Auditable by design
TLS 1.3, mTLS between services, SHA-3-256 hash-chained audit log, DLP integration, NIST 800-88 secure deletion, post-quantum-ready cryptography. Every action is regulator-inspectable.
Hardware-rooted licensing
Hardware root of trust prevents binary tampering and model swapping. Source code escrow with a 90-day-breach trigger means no vendor lock-in: if we ever fail you, you keep the build.
Enterprise integration
Multi-tenant, RBAC, SSO (SAML 2.0, OIDC), LDAP / AD, BYOK provider credentials. Hungarian KAÜ / DÁP / eIDAS shipped natively. ASR add-on for batch and real-time meeting transcription.
Four deployment modes — one codebase
Internet-connected, multi-tenant SaaS.
Customer-managed, allowlisted egress.
Strict allowlist (KAÜ / DÁP / eIDAS).
Zero egress. Fully offline.