Legal
AI Use Disclosure
Last updated August 2026
AI-generated work
chAIn agents use generative models to plan and produce work. Model output can be incomplete, inaccurate, biased, or inappropriate even when it is presented confidently. Your organization remains responsible for reviewing output and configuring approval requirements proportionate to the impact of each task.
Tools and consequential actions
An agent may call only tools permitted by its configuration and the organization's policy. Tool arguments are validated, higher-risk actions can require an approval bound to the exact arguments, and code execution uses an isolated sandbox. These controls reduce risk but do not make unsupervised consequential use appropriate.
Model providers
A configured model provider processes the context needed for a request. chAIn may route among approved providers according to capability, policy, availability, and cost. Provider identities are not exposed in customer-facing agent output, but authorized organization administrators can review the routing and data-classification controls that govern their agents.
Untrusted content
Uploaded documents, retrieved knowledge, webpages, messages, and tool responses are untrusted data, not authority to change system instructions. chAIn scans for prompt injection, secrets, canaries, and personally identifiable information; customers must still minimize sensitive input and verify generated output before relying on it.
Human accountability
Deploying an agent does not transfer legal or professional responsibility to chAIn or to a model. Organizations must tell affected people when AI involvement is material and provide human review, appeal, or correction mechanisms where law or context requires them.