Security and AI governance
Customer lists, pricing, production data, drawings, the know-how that wins you contracts. That is exactly what your team is tempted to paste into a public chatbot. InnateIQ gives them a better place for it: one that is walled off for your company alone, encrypted, governed by your roles, and never used to train anyone's model.
Defense in depth: every request passes every layer
Why not just use a chatbot?
General purpose AI chat is a remarkable tool, and a risky home for company data. When someone pastes a customer list into a personal chat account, it leaves your control: no roles, no audit trail, no say over how it is kept. InnateIQ brings the same intelligence inside a boundary your company owns.
| Pasting into a consumer AI chat app | Working with OTIS in InnateIQ | |
|---|---|---|
| Model training | Depending on the plan and settings, conversations can be used to improve the provider's models. | InnateIQ never uses your data to train models. OTIS runs on enterprise model APIs, not consumer chat apps. |
| Who controls it | Lives in an individual's personal account. When they leave, so does your visibility. | Lives in your company's workspace, under your admins, with your roles. |
| Separation from others | A shared consumer service. Separation is a matter of the provider's account system. | Every customer's data is isolated by row-level security in the database itself, and denied by default. |
| Who can see what | Whoever has the login sees everything in it. | Owner, admin, member and viewer roles, enforced on every screen, every API call and every action OTIS takes. |
| Taking action | Copy, paste, retype into the CRM, and hope nothing was lost on the way. | OTIS writes to your systems of record, and only after a person approves the exact change. |
| Accountability | No company audit trail of what was shared or done. | Key changes are written to an audit log, and every model call is tracked per organization. |
| Shadow AI | Each employee picks a tool. IT finds out later. | One sanctioned place for AI work, configured once and governed centrally. |
Consumer AI products differ by plan, and their policies change. Enterprise tiers of frontier models offer stronger terms; InnateIQ is built to deliver those protections plus the controls a business needs around them.
What protects you today
Your records are separated from every other customer by row-level security inside PostgreSQL itself. If a request ever arrives without a verified company context, the database returns nothing. It fails closed.
InnateIQ runs on Amazon Web Services in the United States, with services inside a private network and data encrypted at rest. The keys that connect your other systems are encrypted again, field by field.
OTIS can look freely within your permissions. Updating a record, sending an email, archiving an account: each one comes back to you in plain language, and nothing happens until someone says yes.
Enterprise sign-in through Auth0 with organization-level separation. The roles your company assigns apply identically to people and to OTIS, and OTIS always acts as the person who asked, never as anyone else.
Each person connects their own inbox, calendar and CRM. Their credentials are invisible to coworkers, and OTIS only gains a capability when that person has connected it, so access never quietly spreads.
Every change to InnateIQ passes automated checks before it can ship: secret scanning, dependency monitoring, and a gate that fails the build if any new data table lacks isolation rules. Rate limits protect every endpoint.
Human in the loop
41
Actions that wait for a yes
Agentic AI is only useful if you can trust what it does on your behalf. Every action that changes a record or reaches outside your company, from a CRM update to a campaign send, is held as a pending decision until a person approves it.
We are integrating a dedicated AI governance platform into InnateIQ. It inspects what is about to reach a model, enforces your privacy policy before anything leaves, and turns compliance evidence into a by-product of the work.
PII shield
Every prompt and document is scanned for 15 categories of personal information first. You decide, category by category, whether each is allowed, redacted, replaced with a secure token, or blocked outright. Blocked requests never leave your environment, and tokens are restored in the answer so your team still gets a useful reply.
Bring your own model keys
Connect your own OpenAI, Anthropic or OpenRouter credentials, so AI usage runs under the agreements your company has already negotiated. Keys are encrypted with AES-256-GCM through AWS Key Management Service, scoped to a single project, and never shared or used as a fallback for anyone else.
Compliance evidence
A tamper-evident evidence ledger for SOC 2: every record hashed, every event appended in sequence, and every package approved by a person before it goes to an auditor. ISO 27001 and HITRUST collection are next on the roadmap.
Spend and usage guardrails
Usage ceilings are reserved before each paid AI or cloud call, not reconciled after it. When a limit is reached, calls stop. Paired with project-scoped API keys that are stored only as hashes and can expire or be revoked at any time.
Passkey sign-in
Passwordless sign-in with passkeys, with verification enforced on every administrative session, so the accounts that can change the most are the hardest to compromise.
Zero-retention routing
We are moving every model request to providers that commit not to train on or retain your prompts, enforced in code on every call rather than left to an account setting.
Questions security teams ask
A full security and AI governance overview is available for your IT and compliance team on request.
No. InnateIQ never uses your data to train models, and OTIS runs on enterprise model APIs rather than consumer chat products. We are also moving all model traffic to providers that commit to retaining nothing, enforced on every request.
No. Every company's data is separated by row-level security in the database. Requests without a verified company context get nothing back, and our test suite attacks this boundary on every change, using the same restricted database role that runs in production.
No. OTIS can read what your role allows. Anything that changes a record or reaches outside the company is shown to you as a pending decision in plain language, and only happens after a person approves it.
On Amazon Web Services in the United States (Ohio region), in an encrypted Amazon Aurora PostgreSQL database, with application services running inside a private network.
Your admins assign owner, admin, member or viewer roles. Those roles are enforced on every screen and API, and OTIS follows exactly the same rules as the person it is working for. Personal items, such as a connected inbox, stay private to their owner.
We are on path toward SOC 2 and HITRUST certification, and the platform is built around the controls those frameworks require. We are happy to walk your team through our current controls in detail.
The closed beta
InnateIQ is in closed beta with a small group of manufacturers. Bring your security questions; we will bring the answers.