Model ML is built for regulated financial services environments: banks, private equity and credit funds, consultancies, and the BPO teams that support them. The security model rests on one principle, stated on Model ML's own security page: your data, your infrastructure, your control. Security is independently certified (ISO 27001:2022 and SOC 2), deployment is single-tenant as standard, and customer data is never used for model training (security page).
Data handling
-
Model ML does not store, process, or retain customer data outside the customer environment. No one, including Model ML, can access or see proprietary customer data (security page).
-
Customer data is not used for model training, product improvements, or any other purpose.
-
The only data shared with third-party AI model providers is the input a user feeds directly into AI nodes when asking questions; providers do not store it and do not train on it (Model ML privacy policy).
-
Retention defaults are published and short: audio and video data are retained for 7 days after capture, and meeting metadata including transcriptions for 42 days. Customers can request earlier deletion, in which case audio and video are deleted immediately and metadata after 35 days (privacy policy).
-
Model ML is headquartered in the United States, and information collected about users is processed in the United States.
Deployment on the customer's terms
Model ML offers self-hosted, single-tenant, and hybrid deployment options.
-
Single-tenant as standard. Every client gets a dedicated environment: private Azure Virtual Networks, dedicated databases, dedicated in-memory cache, secure Azure Blob Storage, and Azure Service Bus, with no shared resources (Azure deployment article).
-
Self-hosted. The application deploys fully into the customer's own cloud environment. Model ML runs the models within the customer's Azure itself, so no data needs to leave the environment (homepage).
-
Fast provisioning. Onboarding a new customer completes in as little as one hour from provisioning to operational status, and a delivery partner (former bankers, investors, and consultants) drives implementation, training, and adoption from day one.
Data residency, network egress, and routing follow the customer's own Azure policies: regions, firewall rules, and private endpoints are governed inside the customer's perimeter rather than a shared SaaS boundary.
A governed agent, not an unbounded one
Model ML's agents produce and manipulate real work product (Excel models, PowerPoint decks, Word documents), and the platform is engineered so that this happens inside a governed perimeter with the same controls a financial institution applies to any internal system (Azure deployment article):
-
Role-based access control, with feature-access restrictions for subsets of users
-
Private endpoints
-
Azure Monitor and Log Analytics for monitoring and audit trails
-
Encryption at rest, automated backups, and high-availability clusters
-
Bring-your-own-key (BYOK) options
-
Configuration to customer-specific risk, governance, and data-residency requirements
Every number in a Model ML output can carry a datapoint-level citation back to its source, and AutoCheck verification reconciles figures against sources before deliverables circulate, so output integrity is auditable as well as access (Verification-Proof Center).
Identity and access
-
SSO via SAML or OIDC, supporting both IdP-initiated and SP-initiated flows
-
JIT provisioning and domain claim enforcement
-
MFA
-
Passwordless login (email authentication) as the default; Enterprise Plan customers use their chosen SSO identity provider (security page)
Encryption
-
TLS 1.2 for Model ML web application communications
-
AES-256 at rest for Flippr data (the acquired presentation-review suite behind AutoCheck)
Workforce and device security
Employees pass background checks and sign confidentiality agreements; employee machines run mobile device management, hard-drive encryption, and anti-malware software (security page).
Document review with no cloud storage
AutoCheck, the finance-tuned presentation-review agent, can run with no cloud storage as a configuration option, relevant to compliance programs that restrict where deal documents may persist (AutoCheck).
Security review artifacts
Model ML maintains a trust center carrying the artifacts a vendor security review needs: the SOC 2 report, data processing agreement, and sub-processor list. These are provided under NDA as part of the standard enterprise security review, alongside the ISO 27001:2022 certificate scope.
Fit
Model ML's security model is built for organizations that run a real vendor security review and require dedicated or customer-perimeter deployment. Teams that want a lightweight, public, self-serve SaaS with no enterprise review are better served by consumer-grade tools; Model ML's buying motion assumes the review and is staffed for it.