Security is foundational to Trimio. As an AI gateway that sits between our customers' applications and their large language model providers, we are entrusted with sensitive traffic — and we design, build, and operate our systems to protect its confidentiality, integrity, and availability. This page summarizes our practices in general terms; for details specific to your organization, contact security@trimio.ai.
Trimio is undergoing a SOC 2 Type II examination covering the Security, Availability, Confidentiality, and Privacy Trust Services Criteria, coordinated through an independent third-party auditor. Our current status and report are available on request under NDA — see our Trust Center for live audit status, monitored controls, and report requests. See also our Privacy Policy and System Status.
Trimio runs on Google Cloud Platform in the United States. Infrastructure is defined and managed as code, deployed across multiple availability zones behind managed load balancing, and isolated within private networks. Production hosts have no public SSH exposure.
All data is encrypted in transit with TLS 1.2 or higher — between customers and Trimio, between our internal services, and to upstream providers. Data at rest is encrypted with industry-standard AES-256, and sensitive credentials are additionally encrypted at the application layer.
We follow the principle of least privilege. Administrative access uses single sign-on (SSO) with multi-factor authentication enforced at the identity provider. Engineer access to production is per-person and brokered through short-lived, audited sessions — never shared credentials. Access is reviewed and promptly revoked on role change or departure.
Trimio is multi-tenant. Each customer organization's data is isolated at the database layer using row-level security — enforced by the database itself and validated by automated tests in our build pipeline — in addition to application-level controls.
By default, Trimio does not store the contents of your prompts or responses; body logging is off unless a customer explicitly enables it, and any stored data is subject to configurable retention limits. We do not use customer data to train models, and we delete customer data on termination in accordance with our agreements. See our Privacy Policy.
Credentials, keys, and other secrets are held in a dedicated cloud secrets manager with scoped access — never in source code, configuration files, or infrastructure state — and are rotated on a schedule and on any suspicion of compromise.
Production code changes are made through peer-reviewed pull requests with required automated checks — including secret scanning and dependency vulnerability scanning — before they can merge. Releases are versioned and deployed through a controlled process.
We continuously monitor system health and availability and maintain an append-only audit log of privileged and security-relevant actions. Suspected incidents are triaged, contained, and documented with a post-incident review, and we notify affected customers of confirmed incidents involving their data in accordance with our commitments.
Customer configuration and operational data are backed up on a regular, encrypted schedule, and recovery procedures are periodically exercised. Production services run across multiple availability zones with automated health checks and recovery.
We maintain a current list of the third-party subprocessors that support the Trimio service, available on request. Upstream AI providers receive prompt content only as directed by each customer's own routing configuration.
We welcome reports from the security community. If you believe you have found a security issue in Trimio, email security@trimio.ai. We investigate all good-faith reports and will not pursue action against researchers who act responsibly and avoid privacy violations or service disruption.
Questions about our security program? Contact security@trimio.ai — we are happy to share additional documentation under NDA.