Arena AI is most effective for providing visitors with recommendations based on your publicly available content. It is intended to be a helpful tool.
We have implemented internal guardrails and processes, both automated and human, that protect against certain types of harmful hallucinations, including but not limited to:

Additionally, we continuously monitor AI responses through random sampling and adversarial testing, among other methods. However, as we rely on a third party language model, any unforeseen hallucinations or biases may reflect the language models’ inherent vulnerabilities. Such incidents do not reflect Arena’s views. When appropriate, Arena will help the Customer understand and improve upon provided Services so that such incidents can be avoided in the future.

Arena AI Services are not intended to replace editorial standards. We recommend that you always validate the accuracy of content generated by Arena AI before relying on it for critical workflows.
​