Security and Trust Center
Your grant strategy, evidence, and funder relationships deserve explicit controls.
AutoGrant handles sensitive organizational context. This page explains what AutoGrant does today, what remains the customer's responsibility, how human approval is enforced, and how dedicated or customer-controlled deployment can support organizations with stronger governance requirements.
Last updated: August 7, 2026
Trust principles
Five commitments this page is built on.
Everything below follows from these principles — including the places where the honest answer is a careful one.
Customer data is not product training data
AutoGrant does not use customer content to train shared models or inform another customer's workspace. Model-provider and subprocessor terms must support the contractual data boundary.
Human approval is a product boundary
No application is submitted without explicit authorization from a named person. Preparation and approval remain separate.
Least necessary access
Each deployment defines who may access production systems and customer content, for which approved purposes, for how long, and what logging and review apply.
Clear ownership and exit
Customers own their data and outputs. Export, retention, and deletion are governed by the agreement and published policy.
Honest compliance posture
AutoGrant states certifications and controls precisely and does not imply compliance that has not been independently established.
Current controls
Controls, stated for the deployment they apply to.
Each control below is described precisely for the deployment model it belongs to — hosted, dedicated, or customer-controlled. Where a control depends on the agreement, the description says so.
Encryption in transit
Traffic to and from AutoGrant uses current supported TLS protections.
Encryption at rest
Application databases and file storage use cloud-provider or customer-managed encryption controls appropriate to the deployment.
Identity and access
Access is authenticated and scoped by role. Enterprise deployments may use customer identity and access controls where supported and agreed.
Environment separation
Hosted customer data is logically separated. Dedicated Enterprise environments provide stronger isolation. Customer-controlled deployments run in the agreed customer environment.
Logging and monitoring
Administrative and security-relevant activity is logged according to the deployment and support model.
Backup and recovery
Backup, recovery objectives, testing, and responsibility are defined by the hosting and Enterprise agreement.
Administrative access
The applicable deployment defines authorized production-access purposes, role boundaries, logging, review, and incident-response procedures.
Model-provider controls
The model connection, retention terms, data use, location, and provider responsibilities must be documented for the applicable deployment.
Secure development
Dependencies are reviewed and patched when an advisory affects the product, and every change builds through a managed deployment pipeline before it reaches production. Where a practice is still being formalized rather than fully established, we say so plainly instead of implying maturity the product has not reached.
Human authorization
Nothing is submitted without named human authorization.
AutoGrant separates:
- Opportunity analysis
- Drafting and preparation
- Factual and program review
- Budget and compliance review
- Final submission authorization
Where AutoGrant supports approved portal or workflow actions, the final step requires explicit authorization. Autonomous final submission is not permitted.
The review record can preserve who reviewed:
- Facts and sources
- Program commitments
- Budget and financial representations
- Compliance and required disclosures
- Final narrative
- Submission authorization
Data use and AI
What is processed, what is never done, and what stays yours to govern.
What AutoGrant may process
Approved organizational sources, public funding information, opportunity documents, drafts, workflow records, reviews, and other customer-authorized content.
What AutoGrant does not do
- Train shared models on customer content
- Use one customer's context to improve another customer's output
- Sell customer grant strategy or documents
- Treat generated facts as verified without human review
- Independently submit an application
Customer responsibility
The customer defines what content may be used, obtains necessary rights and approvals, verifies outputs, follows funder and institutional policy, and prohibits sensitive data not covered by the deployment and agreement.
Hosted deployment
Two hosted models.
AutoGrant Platform
A shared product service with customer data boundaries, standard controls, and guided onboarding.
Dedicated AutoGrant-operated environment
An Enterprise option with customer-specific application and data resources, agreed access, support, release, monitoring, backup, and recovery.
The exact isolation, recovery, identity, and support model must be documented in the agreement.
Customer-controlled deployment
Run AutoGrant where your organization governs the environment.
AutoGrant Enterprise may be deployed in an approved customer cloud or infrastructure when supported by the architecture and agreement.
This can help align the deployment with:
- Customer identity
- Networking and private connectivity
- Logging and monitoring
- Data residency
- Key management
- Retention and backup
- Security operations
- Procurement and vendor-management requirements
Shared responsibility
Customer-environment deployment does not automatically make AutoGrant compliant with every law, policy, certification, or institutional standard.
Assign responsibility for:
- Cloud and network configuration
- Identity and access
- Endpoint and administrator security
- Application configuration
- Model-provider connection
- Data classification and approved use
- Logging and response
- Patch and release management
- Backup and recovery
- User training and workflow governance
Procurement and review
What your review can request.
AutoGrant can support review through current, verified:
Security review should be identified early in the Enterprise process and included as a named workstream.
- Security questionnaires
- Architecture and data-flow documentation
- Subprocessor list
- Data-processing terms
- Accessibility documentation
- Responsible-AI and human-review documentation
- Retention and deletion description
- Incident-notification terms
- Customer-environment responsibility matrix
- Public-sector or higher-education assessment formats where applicable
Current compliance posture
AutoGrant is an evolving product and does not claim certifications or regulated-use compliance it has not earned. Current certification, audit, and deployment status is documented in the Security overview. If your organization requires a specific certification, contractual control, or regulated-data use, raise it at the beginning of the review so we can state clearly whether the requirement is supported today.
Responsible AI
AutoGrant's responsible-AI approach includes:
- Human accountability
- Transparent score reasoning
- Source and assumption visibility
- Customer data boundaries
- Contestable recommendations
- Explicit final approval
- Prohibition on fabricated evidence
- Support for funder or institutional disclosure requirements
- Minimum necessary autonomy
AutoGrant is a signatory of the Fundraising.AI Framework for Responsible & Beneficial AI.
Vulnerability disclosure
Report a security concern
Email security@autogrant.ai with:
AutoGrant will acknowledge good-faith reports according to the published disclosure policy and asks researchers to provide a reasonable opportunity to investigate and remediate before public disclosure.
- Description
- Affected URL or component
- Reproduction steps
- Potential impact
- Safe supporting evidence
- Contact information
FAQ
Security questions, answered precisely.
Request the Security overview
Start your review with current documentation.
Request the current Security overview, architecture information, data-processing terms, and deployment responsibility model. Tell us what your review needs — and where something is not yet supported, we say so.
Give your security and procurement team a precise answer.
Request the current Security overview, architecture information, data-processing terms, and deployment responsibility model.