Chrysent Touchstone

Chrysent Touchstone protects the conversation boundary around AI systems. Bastion prevents assistants from taking harmful actions through tools and infrastructure; Touchstone evaluates the content itself, including prompts, responses, intent, and unsafe interaction patterns.

It can identify prompt injection, jailbreak attempts, malicious inputs, unsafe outputs, illegal usage, PII, and sensitive information. Sensitive values can be redacted before content is sent to a model and restored for the user when the response is returned.

Touchstone helps organizations operate AI where privacy, safety, and content controls are required by policy or law. Paired with Gateway, it makes classification, redaction, blocking, reporting, and review part of the same control plane that tracks usage, cost, audit history, and operational data.