Responsible AI & Model Governance Principles

Last Updated: September 16, 2025

At Hund, we are committed to using generative models responsibly. These technologies, such as large language models (LLMs), can enhance how customers interact with our platform. Our approach focuses on security, privacy, and customer control.

Principles

Transparency and Human Review

Features powered by generative models will always be opt-in. Customers decide when and how to use them. Hund does not allow models to act outside of the access a customer has explicitly authorized. Sensitive actions such as publishing or deletion require specific scopes or permissions and are logged for accountability. When customers connect an external model through MCP, they are responsible for how that model uses the permissions they grant.

Control

Customers remain in control of what models can access. Hund enforces permissions through scoped API keys or role-based access. Whether customers connect models through the Model Context Protocol (MCP) or use hosted features, the scope of access is clear and limited to what has been authorized.

Privacy

Customer data belongs to customers. Hund will never use customer data to train or fine-tune generative models. Data exchanged with a model is processed only to complete the requested action and is not stored or reused.

Security and Reliability

Generative model features will follow the same security and quality standards as the rest of the Hund platform. We validate inputs and outputs to reduce risks such as injection attacks or misuse. Hund is responsible for the safety of the tools and APIs we provide. Customers are responsible for the behavior of external models they connect through MCP.

Fairness

Generative models can reflect bias from their training data. For features Hund provides directly, we will evaluate harmful bias and take steps to mitigate it.

Ongoing Commitment

Hund monitors new standards and regulations, including the EU AI Act and the NIST AI Risk Management Framework. We will review and update this policy to reflect best practices as the technology and rules evolve.

Questions

Contact privacy@hund.io with any questions about our approach to generative models.