R2 Copilot

R2 Copilot targets the 'shadow AI' problem: employees already paste company material into consumer chatbots, and blocking them simply slows the work down. Instead of forcing a new tab or a new habit, R2 installs into existing surfaces — native add-ins for Microsoft 365 (Outlook, Word, Excel, PowerPoint, OneNote, Teams), Google Workspace apps via the Marketplace, a browser extension and web console, and a bot for Telegram. From that single layer, a team can pick the model that suits the task: GPT, Claude, Gemini, Grok, Llama, DeepSeek, or open-weight models hosted on their own infrastructure.

The differentiating technology is the SDNP encryption and routing path. Prompts are sharded and encrypted with AES-256-GCM at the client endpoint, then routed through a zero-knowledge mesh that applies PII/PHI controls before anything exits. The model receives no origin and no identity, which lets regulated buyers use frontier AI while keeping data at rest out of vendor retention windows. The company states the design is covered by US patent 9,998,434 B2 and built on SOC 2 infrastructure.

Governance is the third pillar: role-based access, policy rules, and in-tenant audit logs intended to satisfy auditors and frameworks such as HIPAA, SEC/FINRA and NYDFS obligations, with GovCloud, air-gapped and BYOK options at the enterprise tier. Public plans run from a single-seat Pro tier up to flat-rate Team and Business plans and a quoted Enterprise agreement, so the product scales from individual researchers to sovereign deployments.