Datafold Security Addendum
| Effective | Status |
|---|---|
| October 13, 2026 | Current version |
1. Scope and application
1.1. This Security Addendum forms part of the agreement under which Datafold, Inc. provides Services to Customer (the “Agreement”) when incorporated by that Agreement or the Data Processing Addendum (“DPA”). It supplies the Security Measures incorporated into DPA Annex II. Capitalized terms not defined here have the meanings in the Agreement or, if undefined there, the DPA.
1.2. Datafold maintains the safeguards below (“Security Measures”) for the Platform components it operates and for its personnel, including when performing Professional Services in Customer’s environment. Services under a Statement of Work (“SOW”), including proofs of concept, are covered; free trials are excluded. Section 4 allocates responsibilities by deployment.
1.3. The Agreement and DPA govern confidentiality, permitted processing, incident notification and cooperation, deletion procedures, audit rights, warranties, remedies and liability. Customer-specific security requirements in an Order Form, SOW or partner agreement prevail to the extent stated. Datafold may update this Addendum at https://www.datafold.com/legal/security-addendum, with version history, provided no update materially diminishes the overall protection of Customer Data during the term; the version incorporated governs until so updated.
2. Security program and personnel
2.1. Datafold maintains a written, risk-based security program aligned to the AICPA Trust Services Criteria for Security, Availability and Confidentiality. Management reviews its policies and risks at least annually and assigns responsibility for the program. Datafold assesses vendors before engagement and reviews them at least annually according to risk; vendors accessing Customer Data are bound by written confidentiality, security, incident-notification and deletion obligations appropriate to their access. The DPA governs Subprocessor obligations and changes.
2.2. Datafold requests criminal-record and reference checks for employees before or within 30 days after starting, where lawful. Management evaluates adverse or incomplete results. Contractors with privileged or administrative access to production systems or Customer Data undergo proportionate screening or provide acceptable background evidence. Additional screening required by an Order Form or SOW is completed before access.
2.3. Personnel acknowledge confidentiality obligations and security policies before access and complete security and privacy awareness training at least annually. Access is reviewed on role change, removed promptly when no longer needed, and removed within one business day of termination.
3. Access and devices
3.1. Production and Customer Data access requires documented approval, a business need and least-privilege roles; access is reviewed at least quarterly. Personnel use unique accounts, workforce MFA and MFA-protected privileged access. Administrative access to production control planes Datafold operates uses a VPN or equivalent private path. The Platform supports SAML/OAuth single sign-on, role-based permissions, salted password hashes and password length and complexity requirements for local accounts, and revocable API keys.
3.2. Customer-system credentials are encrypted at the application layer, excluded from source control and logs, and accessible only to components that need them. Encryption keys are stored separately from encrypted data; custody follows Section 4. Key access is restricted and logged; application keys and credentials are rotated on suspected compromise.
3.3. Personnel access Customer Data only through company-managed endpoints or devices issued under Partner-Access, never personal devices. Datafold inventories company-managed endpoints with Customer Data access. Personnel must keep these endpoints patched, screen-locked, and disk-encrypted under Datafold’s endpoint policy. Media and devices are securely erased before reuse or destroyed at disposal. Customer Data is not held on paper or removable media.
4. Deployment Models and Responsibilities
4.1. Each Order Form or SOW identifies the Deployment Model, this Addendum’s version, and any Partner-Access arrangement:
- SaaS: shared, multi-tenant infrastructure hosted and operated by Datafold in its cloud account.
- Single Tenant Datafold-hosted: a dedicated Platform instance hosted and operated by Datafold in its cloud account.
- Single Tenant Customer-hosted (Datafold-operated application): the Platform operates in Customer’s cloud account or environment on Customer-operated infrastructure; Datafold operates the application through Customer-granted access.
4.2. “Datafold Cloud” and “multi-tenant SaaS” mean SaaS; “Dedicated Cloud” and “Single Tenant” in Datafold’s account mean Single Tenant Datafold-hosted; “Customer-Managed,” “Customer VPC,” “self-hosted,” “on-premises” and “Single Tenant” in Customer’s account mean Single Tenant Customer-hosted (Datafold-operated application). These aliases and other descriptive deployment names map to a model only where the described hosting and operating responsibilities match that model above. In Customer-hosted deployments, Customer operates the infrastructure and Datafold operates the Platform application. “Partner-access model” and “partner-access arrangement” mean the Partner-Access overlay below. Region selection follows the DPA and Order Form or SOW. SaaS alone is available for free trials.
4.3. The named party performs and is responsible for each control. Controls not listed belong to Datafold for its Platform and to Customer for its own environment and users. “Infrastructure” includes cloud account, identity, network, compute, storage, patching, configuration scanning, infrastructure logging/monitoring, ingress TLS/DNS, edge protection, model-provider egress, infrastructure recovery/testing and cloud-provider physical security.
| Control | SaaS | Single Tenant Datafold-hosted | Single Tenant Customer-hosted (Datafold-operated application) |
|---|---|---|---|
| Infrastructure | Datafold | Datafold | Customer |
| Platform deployment, releases, data-store encryption and application logs/monitoring | Datafold | Datafold | Datafold |
| Cloud KMS keys for Platform stores, disks and backups | Datafold | Datafold, or Customer-managed keys where the Order Form or SOW so provides; Datafold configures their use | Customer |
| Platform data-store backups | Datafold | Datafold | Customer, in its account under its keys |
| Protection of credentials the Platform uses to connect to Customer’s data sources and targets | Datafold | Datafold | Customer provides connection credentials; Datafold protects credentials handled by the Platform under Section 3.2 |
| Release automation, remote access and application-log retention | Datafold | Datafold | Customer approves releases and provides access; Datafold runs automation and configures the retention period specified in the Order Form or SOW |
4.4. Datafold remains responsible for its personnel, managed devices and issued credentials in every model. At the end of the Services, Datafold deletes Customer-specific deployment infrastructure in its account in accordance with the Agreement’s and DPA’s export, retention and deletion periods. Customer deletes deployment infrastructure in its account in accordance with those periods unless the SOW provides otherwise.
4.5. Customer promptly reports suspected compromise of credentials or systems affecting the Services and keeps its security contacts current.
4.6. Partner-Access overlays any model when personnel use accounts, devices, or tooling issued by Customer, its Affiliate, its customer or a partner. Customer is responsible for the security and applicable acceptable-use, export-control, and endpoint policies of what it or the issuer supplies. Datafold personnel follow those policies, use issued resources only as issued and for the Services, use only issuer-authorized AI tools where required, and notify Customer when access is no longer needed. Datafold remains responsible for its conduct, including while using issuer-provided resources.
5. Encryption, networks and data flows
5.1. For controls allocated to Datafold, Platform data stores, volumes, object storage and backups use AES-256 or stronger encryption at rest. Provider-managed keys rotate under provider mechanisms and Datafold’s cryptography policy. Customer Data uses TLS 1.2 or higher between Customer’s users/systems and the Platform, between the Platform and Model Providers, and when logs or telemetry are sent to operations Subprocessors.
5.2. Datafold-hosted workloads use private subnets, managed load balancers, restricted component access, and cloud-provider denial-of-service protection. Datafold-operated load balancers disable legacy TLS protocols and use modern cipher suites. Network rules are reviewed on change and at least annually. Customer-hosted deployments use Customer’s network controls and the deployment guide’s component segmentation. Datafold uses dedicated, least-privilege credentials for Customer systems, revocable by Customer on notice subject to the SOW.
5.3. Customer-hosted deployments process source data in Customer’s environment. Only operational telemetry/logs may be sent to operations Subprocessors on the Subprocessor List identified in the Agreement. Requests to Model Providers may leave Customer’s environment as permitted by the Agreement. Any Customer Data in these outbound flows remains protected under the Agreement and DPA. Logs are designed to exclude data values but may contain identifiers or error-related query fragments; these are protected as Customer Data. Customer may require telemetry and logs to remain in its environment through the Order Form or SOW.
5.4. Tenant isolation. SaaS checks tenant identifiers to restrict access to each customer’s data.
6. Development, vulnerabilities and testing
6.1. Datafold uses peer review, automated testing and security scanning before merge, separate development/staging/production environments with distinct credentials, and additional design review and testing for security-control changes. Production changes require approval and rollback procedures; emergency changes are documented and reviewed afterward. Customer Data is not copied into non-production environments except with documented executive approval, minimization where feasible and production-standard protection. Customer code in Professional Services repositories receives the same access and logging safeguards.
6.2. For infrastructure and application components operated by Datafold, Datafold continuously scans dependencies and container images, scans cloud configurations against recognized benchmarks, performs external production vulnerability scans at least quarterly, and internal scans against production-mirroring environments. Findings are tracked and prioritized by CVSS severity and exploitability. Remediation targets for those infrastructure and application components, measured from severity assessment, are: Critical 7 days; High 14 days; Medium 30 days; Low 90 days. OS and Platform patches follow these targets. Unavailable or disruptive fixes require documented exceptions and compensating controls.
6.3. Datafold notifies Customer through the Trust Portal or directly of confirmed Critical/High findings affecting the Services that exceed those targets, including compensating controls and an expected remediation date. Datafold identifies Critical/High security fixes in releases and promptly notifies Customer of Critical Platform vulnerabilities requiring customer action.
6.4. An independent third party tests the SaaS production application and exposed infrastructure at least annually. Findings follow the same remediation targets; Critical/High findings are retested before closure. Customer testing of Datafold-operated deployments requires written rules of engagement. Customer shares the results with Datafold; confirmed findings follow these targets. Customer controls testing of its infrastructure, subject to the Agreement. Security researchers may report vulnerabilities through the security-notice contact in the Order Form.
7. Monitoring and incident response
7.1. Authentication, privileged actions, configuration changes and errors are logged. Centralized, tamper-protected logs and monitoring alert on-call personnel to anomalous privileged activity, configuration drift and unusual data egress. Datafold maintains an incident response plan, reviewed annually, covering escalation, containment, recovery, evidence and communications; alerts reach on-call personnel 24x7. Notification, investigation cooperation and root-cause reporting follow the Agreement and DPA.
7.2. Logs held by Datafold are retained for 12 months and then deleted, unless an Order Form or SOW specifies another period for a dedicated deployment. Logs in Customer’s account follow Customer’s configuration. Customer Data fragments in logs are used only for security investigation and troubleshooting.
8. Recovery, backups and working data
8.1. Datafold maintains and annually tests a business continuity/disaster recovery plan and monitors operated deployments for capacity, latency and errors. Where Section 4 allocates infrastructure recovery and backups to Datafold, regional-loss objectives are 24-hour RPO and 8-hour RTO, unless the Order Form states otherwise. Recovery and backup restoration are tested annually; Datafold provides Platform redeployment support in every model.
8.2. Where Section 4 allocates backups to Datafold, Datafold makes daily backups encrypted at rest and in transit, promptly copied to a second region or another region specified in the Order Form or SOW. Primary-database backups have a rolling 7-day retention; other Platform backups are retained no longer than 35 days, then overwritten or deleted. Dedicated Datafold-hosted backups are otherwise as the Order Form provides. Customer-hosted backups remain Customer’s responsibility under Section 4.
8.3. Retention, export and deletion procedures follow the DPA and SOW. Temporary tables in Customer’s systems are scheduled for deletion within 72 hours after a run and removed by the cleanup process; Customer may drop them at any time. Working data transferred for comparisons that cannot run by hash is deleted when the comparison completes.
9. AI and data minimization
9.1. The Agreement and DPA govern AI enablement, Model Provider selection, Customer-Provided Model Endpoints, no-training obligations and provider terms. Personnel use only approved, company-provisioned AI tooling with Customer Data, never consumer, free-tier or personal AI services; Partner-Access restrictions also apply. Customer-only model training or tuning requires written opt-in in the SOW.
9.2. Customer controls source access and may exclude columns. Where supported, warehouse OAuth connections use the connecting user’s permissions, including masking and row/column restrictions. Professional Services personnel access only what Customer grants. Datafold recommends a Customer-Provided Model Endpoint in Customer’s cloud account for Customer-hosted or PHI-designated deployments.
10. Assurance and supporting documentation
10.1. Datafold maintains an annually examined SOC 2 Type II attestation covering Security, Availability and Confidentiality, or an equivalent successor. Its scope covers SaaS production; coverage of common application operating processes is as the report states. Dedicated Datafold-hosted deployments are not separately described, and Customer-hosted infrastructure is outside the report. Cloud infrastructure providers used by Datafold hold SOC 2 and ISO/IEC 27001 attestations for physical/environmental controls, reviewed before use and annually.
10.2. The Trust Portal (https://security.datafold.com/) provides reports, policies and security explanations under the DPA’s assurance procedure. These supporting materials do not independently amend the Agreement or incorporate every internal policy detail as a contractual commitment. Security notices use the Order Form contacts. PHI-specific requirements remain subject to the applicable BAA.