1. Purpose of this disclosure
Keeptrusts believes organizations should be transparent about how they develop, deploy, and govern AI systems. This page describes how AI is involved in the Keeptrusts platform and services, consistent with the transparency objectives of the EU AI Act (Regulation 2024/1689).
2. How Keeptrusts uses AI
Keeptrusts operates as an infrastructure and governance layer between applications and AI providers. The platform routes, evaluates, logs, and enforces policies on LLM requests and responses.
The gateway routes traffic between customer applications and configured AI providers. Policy evaluation can combine deterministic rules with configured model- or service-assisted checks, including LLM judges, embeddings, and moderation providers.
3. Risk classification under the EU AI Act
EU AI Act roles and risk classifications depend on the intended use, the system in which the gateway is deployed, and the responsibilities of each operator in that value chain. A gateway deployment is not assigned one blanket risk classification by this page.
Customers should assess the classification and obligations that apply to their complete AI system, including downstream models, configured policy services, affected persons, and the decisions supported by that system.
4. Policy enforcement
Policy enforcement can use deterministic conditions such as keyword matching, regular expressions, and structured metadata checks, as well as configured model- or service-assisted checks such as classification thresholds, embeddings, moderation services, and LLM judges.
Customers configure which policies apply to their traffic. Keeptrusts provides built-in policy templates and allows customers to define custom rules.
5. Supporting customer transparency obligations
Keeptrusts provides disclaimer-injection policies and AI-generated content labeling capabilities that customers can configure as part of their transparency controls, including for obligations addressed by Art. 50 of the EU AI Act.
Customers are responsible for determining which Art. 50 obligations apply to their use cases and for configuring, testing, and operating the relevant controls.
6. Data handling
Keeptrusts processes AI traffic on behalf of customers. Prompts, responses, and metadata are handled according to customer configuration, applicable retention settings, and the terms of the governing customer agreement.
Keeptrusts does not use customer AI traffic to train models or improve third-party AI systems. Audit records are retained for compliance and operational purposes as configured by the customer.
7. Limitations
AI governance systems are probabilistic. While Keeptrusts improves oversight, policy consistency, and auditability, no system can guarantee that every unsafe or non-compliant output will be detected in all circumstances.
Customers retain responsibility for the appropriate use of AI within their organization and for evaluating whether governance controls meet their regulatory and ethical requirements.
8. Contact
For questions about this disclosure, contact legal@keeptrusts.com.