Treasurium builds and runs custom AI agents for our customers. This policy explains, in plain language, how we use AI, what customers and website visitors can expect, and how we work to keep AI use safe and accountable.
1. Use of third-party AI models
Our services rely on third-party foundation-model providers. We select providers based on capability, reliability, and their commitments around security and data handling. We do not publicly name specific vendors on this page; a current list is available to customers under NDA on request.
2. Human oversight
Humans remain in the loop for the design, deployment, monitoring, and material updates of every agent we build. Customers retain decision-making authority for actions their agents take on their behalf. Agents are configured with guardrails, approval steps, and logging appropriate to the task.
3. Customer data and training
- We do not use customer data to train foundation models. Where we send customer content to a model provider, we do so under agreements that prohibit that provider from training their base models on customer data.
- Customer data may be used to configure, evaluate, and improve the agents we build for that customer, under the terms of the customer's agreement with us.
- We apply access controls and retention limits proportionate to the sensitivity of the data. See our Data Security Statement.
4. Accuracy and limitations
AI systems can produce output that is inaccurate, incomplete, out of date, biased, or otherwise unsuitable. Output should be reviewed by a qualified person before being relied on for consequential decisions. Our services are not a substitute for professional financial, legal, tax, medical, or other expert advice.
5. Prohibited and high-risk uses
We do not build or operate agents intended to deceive, manipulate, or harm people. We apply additional scrutiny to any use case involving credit, employment, healthcare, immigration, biometrics, safety, or other high-risk domains, and we may decline work that we consider unsafe or non-compliant. See our Acceptable Use Policy.
6. Transparency and disclosure
Where agents built by Treasurium interact with third parties on a customer's behalf (for example, in support conversations or outbound calls), we work with the customer to ensure appropriate disclosure that an automated system is being used, in line with applicable law.
7. Incident handling and feedback
If you believe an agent we built has produced harmful, unsafe, or materially inaccurate output, please tell us. We investigate reports and, where appropriate, adjust configuration, guardrails, or deployment.
8. Contact
AI-related questions or concerns: contact@treasurium.ai.