Certainly! Two perspectives to consider. For one, we hold the keys to the model, much as Anthropic and OpenAI do to there’s. We have guardrails in place, we actively monitor and detect trust and safety violations, and realistically due to the cost there’s far less usage compared
3 0 12