Security & Trust

    Last updated: August 2026

    NovusAI is built on a simple security premise: your data should never have to leave your environment to benefit from AI. Here is exactly how that works, and how to reach us about security.

    Two deployment models — check which one you are evaluating.

    Most of this page describes the self-hosted enterprise deployment, which runs entirely in your own cloud tenant. The hosted service at novusai.org is a multi-tenant service that runs on infrastructure we operate, and its data handling is materially different — it is described in its own section below. Where a control applies to only one model, we say so.

    Self-hosted enterprise deployment

    NovusAI is delivered as signed container images that you deploy inside your own Azure tenant. Every component — the agents, the admin portal, the ServiceNow integration layer, and all data stores — runs on infrastructure you own and control.

    Language-model inference runs on model endpoints you choose and operate: your Azure OpenAI deployments, Azure AI Foundry models, or models you self-host. NovusAI connects to them; it never routes your data through models or infrastructure we operate.

    The data boundary

    Your ServiceNow data, conversations, prompts, credentials, and logs remain in your tenant at all times.

    Shad Corporation Inc. has no inbound access to your deployment, holds none of your credentials, and receives no telemetry. Software updates are delivered outbound-only: you pull signed releases from our registry, and nothing is installed until your administrator approves it.

    The platform requires no connection to us to run. If NovusAI (the company) went offline tomorrow, your deployment would keep working.

    The hosted service at novusai.org

    The hosted service is a multi-tenant application operated by us on Amazon Web Services. It is intended for evaluation and individual use, and it does not carry the data boundary described above. If your organization requires that data never leave your environment, the self-hosted deployment is the option to evaluate.

    • We hold the ServiceNow credentials you enter. Connecting an instance stores its URL, username, password, and OAuth client secret in our database so the agent can act on your behalf. Connect a developer or other non-production instance — never a production one.
    • Your conversations and the agent’s actions are stored by us, including the prompts you send, the tools the agent ran, and the records it read or changed.
    • Model inference runs on Anthropic’s API under our account, not on endpoints you control. Your prompts and the instance data the agent reads are sent to that API to produce a response.
    • The agent can modify the instance you connect. Writes are enabled for connected instances, which is why a throwaway developer instance is the only appropriate thing to connect.
    • Do not put regulated data in it. The hosted service is not appropriate for PHI, payment card data, or anything else your compliance program governs. See the regulated-data section below.

    Regulated data and HIPAA

    Self-hosted deployment. Because it runs entirely in your environment, we do not access, store, process, or transmit protected health information (PHI) or any other customer data. Regulated data stays within your tenant, covered by the controls and agreements you already maintain with your cloud and model providers.

    For most healthcare customers this means NovusAI operates as a software supplier rather than a data processor or HIPAA business associate — a determination your compliance team can validate against the architecture documentation we provide. We are glad to walk your privacy and security reviewers through the data-flow analysis.

    Hosted service. The analysis above does not apply. We operate the infrastructure and hold your data, and we do not currently offer a business associate agreement — so the hosted service must not be used with PHI or other regulated data.

    Support boundary (both models). Our support processes are designed so that PHI is never needed and should never be shared with us. Please do not include patient information in support requests; diagnostic information is reviewed with your team and, for self-hosted deployments, remains in your environment.

    Product security architecture

    The controls below describe the self-hosted enterprise deployment.

    Least privilege, enforced twice

    Each agent tier (End User, Analyst, Developer) is constrained by the tools it is given and, independently, by a dedicated ServiceNow service account whose ACLs cap what it can ever do. The read-only agent’s account holds no write permission at all; the Developer agent’s credentials work only against sub-production instances.

    Identity & access

    Users authenticate through your Microsoft Entra ID single sign-on. Your Entra groups determine who can use which agent tier. The admin portal has its own role-based access control, separating administrators from auditors.

    Encryption & secrets

    All traffic is encrypted in transit (TLS 1.2+). Data at rest is encrypted with keys you manage in your Azure Key Vault, which also holds all credentials — ServiceNow OAuth secrets and model API keys never leave your tenant and are never shared with us.

    Auditability

    Every agent action — who asked, which agent, which tool, what inputs and outputs, when — is written to an append-only audit log in your environment, exportable to your SIEM. Optional human-approval gates can require confirmation before any state-changing action.

    Release integrity

    Agents can only load tools and skills from a signed, versioned registry — and only versions your administrator has approved.

    • Every tool and skill ships with its own automated test suite and must pass it, plus a full regression run of everything previously released, before it can be published.
    • All releases are cryptographically signed; your deployment refuses unsigned or tampered artifacts.
    • A software bill of materials (SBOM) accompanies each release, and dependencies are scanned continuously in our build pipeline.
    • Updates activate only after your administrator approves them, environment by environment, with one-click rollback to the previous pinned version.

    Reporting a vulnerability

    If you believe you have found a security vulnerability in NovusAI or this website, please tell us — we want to know.

    Email security@novusai.org with a description of the issue and steps to reproduce it. We acknowledge reports within one business day and will keep you informed as we investigate and remediate.

    We ask researchers to practise coordinated disclosure — give us a reasonable window to fix an issue before publishing — and we commit in return to not pursuing legal action against good-faith research.

    Business continuity

    We know choosing a young vendor is a risk calculation. Self-hosting is our structural answer: the software runs in your tenant, on your infrastructure, and does not depend on our availability to function.

    For enterprise agreements we additionally offer source-code escrow, so that in the unlikely event Shad Corporation Inc. ceases operations, your organization retains the means to maintain the software you rely on.

    Assurance for your security review

    We support customer security reviews directly and expect to complete your vendor questionnaire as part of any evaluation.

    • Technical architecture documentation, including full data-flow diagrams
    • SBOM for the current release
    • Deployment and hardening guide for your platform team
    • Security review sessions with your InfoSec team during evaluation

    Request any of these at security@novusai.org.