Customer Shared Responsibility Policy

1. Purpose

Trussed AI is delivered to customers as software that customers install, configure, and operate within their own environments. Trussed AI does not operate production infrastructure on customers' behalf and does not process customer data in a Trussed-operated service. This deployment model distributes responsibility for the security, availability, integrity, and compliance of any given Trussed AI deployment between Trussed AI and the customer.

This policy defines the division of responsibility between Trussed AI and its customers. It is the authoritative Trussed AI reference for:

• Customers seeking to understand what Trussed AI is accountable for and what remains their responsibility as they plan, deploy, and operate Trussed AI software.
• Customers performing third-party risk assessments, compliance reviews (for example, SOC 2, HIPAA, NAIC Model Bulletin, NYDFS Circular Letter No. 7, EU AI Act), and audit preparation in which Trussed AI is in scope.
• Trussed AI Personnel responding to customer security inquiries, support requests, and contractual negotiations.
• Auditors and regulators examining the security of the Trussed AI program.

This policy does not create, modify, limit, or expand any rights or obligations under any customer agreement between Trussed AI and a customer. Where this policy and a customer agreement address the same subject, the customer agreement controls.

2. Scope

This policy applies to:

• All Trussed AI employees, cofounders, contractors, consultants, and interns (collectively, "Personnel") in their interactions with customers regarding the security of Trussed AI software and its deployment.
• All customers who install, configure, or operate Trussed AI software, whether under an active commercial agreement, an evaluation agreement, a pilot, or another arrangement that permits use of Trussed AI software.

"Trussed AI software" in this policy includes the Trussed AI product, any components distributed as part of it, any documentation accompanying it, and any patches or updates Trussed AI provides to customers.

3. Deployment Model

3.1 How Trussed AI Is Deployed

Trussed AI software is installed and run in infrastructure owned, rented, or controlled by the customer. Deployments may be in public cloud accounts the customer controls, in private cloud environments, in on-premises data centers, or in hybrid configurations that span more than one of these. The customer chooses the deployment topology that fits its architecture, regulatory posture, and operational preferences.

Trussed AI does not operate infrastructure on customers' behalf. Trussed AI does not host a production service that processes customer prompts, responses, or data, does not hold customer workload data, and does not route customer traffic through Trussed AI-operated systems.

3.2 Implications for Security Responsibility

Because Trussed AI does not operate the infrastructure on which Trussed AI software runs, the customer is responsible for the security of that infrastructure and for the secure operation of the software within it. Trussed AI is responsible for the design, development, and delivery of the software itself, for the features the software provides, for the documentation describing how the software is intended to be deployed and operated, and for the security of Trussed AI's own corporate systems that produce and deliver the software.

The matrix in Section 4 summarizes this division. Sections 5 and 6 describe Trussed AI's and the customer's responsibilities in more detail.

4. Responsibility Matrix

The following matrix summarizes the division of responsibility between Trussed AI and the customer. Rows where both columns are marked represent activities in which both parties contribute to the outcome; Section 7 describes how these are coordinated.

The matrix is a summary. Sections 5, 6, and 7 describe each domain in detail and are authoritative where they elaborate on the matrix.

5. Trussed AI Responsibilities

5.1 Software Development and Supply Chain

Trussed AI is responsible for the secure design, development, and production of the software it ships. This includes:

• Application of secure development practices to the Trussed AI codebase, including peer review, automated testing, static analysis, dependency scanning, and secret scanning, as described in the Software Development and Change Management Policy.
• Security of the source control system, build pipeline, and release infrastructure used to produce Trussed AI software.
• Protection of signing keys and release artifacts from unauthorized use or modification.
• Maintenance of a software bill of materials for shipped releases, available to customers on reasonable request.

5.2 Shipped Software Features and Security Capabilities

Trussed AI is responsible for the security capabilities that the shipped software provides. These include, where applicable to the release:

• Authentication and authorization primitives for use by operators of the deployed software.
• Support for encryption of data in transit between components of the software.
• Generation of logs and audit records that customers can retain, review, and forward to their own systems.
• Controls for prompt and response inspection, policy enforcement, and data filtering as described in the product documentation.
• Configuration options that allow customers to align the behavior of the deployed software with their regulatory and policy requirements.

Trussed AI documents these capabilities and their intended use in the product documentation. Configuring and operating the capabilities within a given deployment is the customer's responsibility.

5.3 Identification of Vulnerabilities and Release of Patches

Trussed AI is responsible for identifying security vulnerabilities in the software it ships, for tracking them to resolution, and for releasing patches in a timely manner. The Vulnerability Management Policy describes Trussed AI's internal process. Trussed AI commits to:

• Receiving vulnerability reports through the channels described in Section 7.1.
• Triaging reports and acknowledging receipt within the timeframes in the Incident Management Plan.
• Developing fixes and releasing them to customers commensurate with the severity of the vulnerability.
• Communicating with affected customers in time for them to take protective action in their own environments.

5.4 Documentation

Trussed AI is responsible for producing and maintaining documentation that describes how the software is intended to be deployed, configured, and operated securely. Documentation includes administrator guides, configuration references, security hardening guidance, and release notes. Trussed AI is responsible for the accuracy and currency of the documentation it publishes. Customers rely on the documentation to configure their deployments appropriately.

5.5 Corporate Information Security

Trussed AI is responsible for the security of Trussed AI's own corporate environment, including the systems that develop, build, sign, and ship the software, the systems that support the business (email, collaboration, finance, HR), and the devices and accounts Personnel use to perform company work. The Information Security Policy and its subordinate policies describe this program in detail.

5.6 Handling of Customer Information Held by Trussed AI

While Trussed AI does not operate a production service that processes customer workload data, Trussed AI does receive and hold certain customer information for sales, support, and contractual purposes. Examples include customer contact information, purchase records, architecture diagrams shared during onboarding, and logs or files shared during debugging of customer-reported issues.

Trussed AI is responsible for protecting this information in accordance with the Information Security Policy, the Access Control Policy, and applicable law. Customer information shared with Trussed AI for support purposes is retained only as long as reasonably necessary for the purpose for which it was shared and is deleted, returned, or anonymized when no longer needed. Customers minimize the information shared for support purposes by avoiding the inclusion of customer data (the data their Trussed AI deployment processes) in support artifacts unless such inclusion is genuinely necessary and consented to.

5.7 Support for Customer Investigations

When a customer investigates a security incident in its own environment that involves Trussed AI software, Trussed AI provides support within the scope of the applicable customer agreement. Support may include clarifying how the software behaves in specific conditions, examining customer-provided artifacts, and identifying whether a vulnerability in Trussed AI software contributed to the incident. Trussed AI does not conduct forensic investigations in customer environments.

6. Customer Responsibilities

Customers are responsible for the security, availability, and appropriate operation of their Trussed AI deployments. This includes the entire environment in which Trussed AI software runs and all data that flows through it. Specific customer responsibilities include, without limitation:

6.1 Deployment Infrastructure

The customer provides and secures the infrastructure on which Trussed AI software runs, including cloud accounts, on-premises hardware, virtualization and container platforms, and the networks connecting them. The customer is responsible for the security configuration of this infrastructure, including account access, resource isolation, vulnerability management, and secure configuration of platform services.

6.2 Operating System and Runtime

The customer is responsible for the operating system, container runtime, and any platform software on which Trussed AI software executes. This includes hardening the operating system, applying security patches to the operating system and runtime, and managing the configuration of platform services in accordance with the customer's own hardening standards and the Trussed AI product documentation.

6.3 Network Security

The customer is responsible for the network architecture surrounding the deployed software, including segmentation between the Trussed AI software and other systems, ingress and egress controls, connectivity to model providers and applications that the deployment is intended to serve, and the security of any network links used. Where the deployment spans multiple environments or geographies, the customer is responsible for the security of the connectivity between them.

6.4 Access Control and Identity

The customer is responsible for determining who may administer the deployed Trussed AI software, who may view its data and logs, and who may change its policy configuration. The customer operates its own identity provider, if any, and integrates it with Trussed AI software using the mechanisms the software provides. The customer is responsible for enforcing its own authentication and authorization policies (including multi-factor authentication, session management, and password policies) and for the lifecycle of users (provisioning, role assignment, deprovisioning).

6.5 Data Protection

The customer is responsible for the data that flows through the deployed Trussed AI software. This includes the classification of that data, the determination of what data may traverse the deployment, handling requirements for any regulated or restricted data, retention and deletion of data and logs, and the consequences of any disclosure or loss of data from the customer environment. Trussed AI software provides capabilities that support the customer's data handling decisions; the customer is responsible for configuring and operating those capabilities to achieve the customer's intended outcomes.

6.6 Encryption and Key Management

The customer is responsible for enabling, configuring, and operating encryption of data at rest and in transit within its environment. Trussed AI software supports common encryption mechanisms and can be configured to require them; the customer must select appropriate options, obtain and install certificates, and manage the lifecycle of cryptographic keys. Where the customer integrates a key management service, the customer operates that service and controls access to its keys.

6.7 Logging, Monitoring, and Detection

Trussed AI software generates logs and audit records of its operation. The customer is responsible for retaining these logs in accordance with the customer's policies and applicable law, for forwarding them to the customer's logging and monitoring systems, for defining alerting thresholds and rules, and for acting on alerts generated. Detection of anomalies, security events, and operational problems within the customer environment is the customer's responsibility.

6.8 Backup, Disaster Recovery, and Business Continuity

The customer is responsible for backing up the configuration and state of its Trussed AI deployment, for establishing recovery procedures appropriate to its availability requirements, and for the design and exercise of its business continuity plans. Availability objectives for the deployed software are the customer's to define and to achieve; Trussed AI does not make availability commitments for software the customer operates.

6.9 Patch Management and Version Control

Trussed AI issues patches and updates to address security vulnerabilities, fix defects, and introduce new capabilities. The customer is responsible for tracking the versions of Trussed AI software it operates, for subscribing to the release communications Trussed AI provides, and for applying patches and updates in accordance with the customer's change management practices. Timing of patch application, coordination with other changes, and validation of patches in the customer environment are customer responsibilities.

6.10 Secure Configuration

Trussed AI software is shipped with secure defaults and with documentation describing secure configuration. The customer is responsible for configuring the software in accordance with the documentation and with the customer's own security and compliance requirements. Deviations from documented secure configurations, including disabling security features for operational convenience, are the customer's responsibility.

6.11 User Management

Users of the deployed Trussed AI software are the customer's users (the customer's employees, contractors, and authorized third parties, and the applications operated under the customer's control that consume or provide data to the software). The customer is responsible for these users throughout their lifecycle: background checks where applicable, training, provisioning, ongoing access review, and deprovisioning.

6.12 Regulatory Compliance

The customer is responsible for the customer's own compliance with laws, regulations, and contractual commitments applicable to the customer and to its use of Trussed AI software. This includes sector-specific requirements (HIPAA, NAIC Model Bulletin, NYDFS Circular Letter No. 7, FDA guidance for software as a medical device), emerging AI-specific regulation (EU AI Act, state AI legislation), and general requirements (state data breach laws, GDPR or other international privacy laws where applicable).

Trussed AI software provides capabilities that support many of these regimes, including policy enforcement, data filtering, and audit evidence generation; the customer configures and operates these capabilities as part of its compliance program. Trussed AI does not determine, warrant, or represent that any given deployment satisfies any particular legal or regulatory requirement.

6.13 Incident Response Within the Customer Environment

Incidents that occur within the customer's environment are the customer's to detect, investigate, respond to, and communicate. This includes unauthorized access to the Trussed AI deployment, disclosure of data the deployment processes, operational failures of the customer's infrastructure, and any regulatory notification obligations the incident triggers. The customer maintains its own incident response capability, separate from Trussed AI's.

Where an incident in the customer environment may involve a vulnerability or malfunction in Trussed AI software, the customer notifies Trussed AI through the channels in Section 7 so that coordinated response may occur.

7. Coordinated Activities

Some activities require contribution from both Trussed AI and the customer. This section describes how Trussed AI expects these activities to proceed.

7.1 Vulnerability Disclosure

Trussed AI accepts vulnerability reports from customers and from external security researchers at hello@trussed.ai. Reporters are encouraged to include sufficient detail to reproduce the issue, an assessment of potential impact, and any relevant context.

On receipt of a report, Trussed AI triages it in accordance with the Incident Management Plan, develops a remediation plan, and communicates with affected customers. Trussed AI asks reporters to refrain from public disclosure until a fix is available or a reasonable period has elapsed. Trussed AI does not pursue legal action against good-faith security research conducted in accordance with this policy.

7.2 Patch Coordination

When Trussed AI issues a patch for a security vulnerability, customers known to be affected are notified through the communication channels identified in their customer agreement and at hello@trussed.ai. Notifications describe the affected versions, the nature and severity of the vulnerability, available mitigations, and the timeline within which Trussed AI recommends applying the patch. Timing of actual patch application is the customer's decision, subject to the terms of any applicable agreement.

7.3 Incident Coordination

Incidents that implicate both parties, including supply chain incidents affecting shipped software and security vulnerabilities exploited in the customer environment, are coordinated between the customer's incident response function and Trussed AI's Security Officer. Each party leads the portions of the response within its own responsibility and shares information with the other party to the extent permitted by law and the applicable customer agreement. Communications with third parties (regulators, law enforcement, media) are coordinated to the extent practical.

7.4 Security Inquiries and Customer Assessments

Trussed AI responds to customer security inquiries, questionnaires, and due diligence requests as part of the sales and renewal process and upon reasonable request during the term of a customer agreement. Requests are sent to hello@trussed.ai or routed through the customer's account contact. Trussed AI provides summaries of its security program, copies of relevant policies and third-party attestations where available, and responses to specific questions.

Customers performing on-site assessments, penetration tests, or security evaluations that involve Trussed AI systems coordinate the scope and timing with Trussed AI in advance.

8. Assumptions About the Customer Environment

The security properties of a Trussed AI deployment depend on the customer environment in which the software runs. Trussed AI's security claims and documentation assume, and the customer's responsibilities in Section 6 imply, that the customer environment has at minimum the following properties:

• Access to deployment infrastructure is restricted to authorized personnel of the customer.
• The underlying operating system, container runtime, and platform services are kept current with security patches.
• Network paths between the Trussed AI deployment and other systems are protected in accordance with the customer's security architecture.
• The identity provider, secrets management, logging, and monitoring systems the deployment integrates with are operated securely by the customer.
• Personnel with administrative access to the deployment are trained, authorized, and accountable.
• Data handled by the deployment is classified and handled in accordance with the customer's own data protection program.

Where the customer environment does not satisfy these assumptions, the security properties of the Trussed AI deployment cannot be assured regardless of the software's own characteristics. Where a customer believes one of these assumptions is impractical in its environment, it is encouraged to discuss alternatives with Trussed AI.

9. Changes to This Policy

Trussed AI may revise this policy from time to time to reflect changes in its deployment model, its software, applicable law, or industry practice. Material changes are communicated to active customers through the communication channels identified in their customer agreements. The most current version of this policy is available on the Trussed AI website and from hello@trussed.ai on request. Revisions are recorded in the revision history below.

Nothing in this policy supersedes the terms of any customer agreement. Where this policy and a customer agreement differ, the customer agreement controls.

10. Policy Review

This policy is reviewed at least annually by the Security Officer and updated as needed. Material changes are communicated to Personnel and to active customers, and are recorded in the revision history below.

11. Revision History