Chrysent Touchstone

The problem this solves

Prompts, responses, and tool inputs can leak PII, accept prompt injection, or produce content your organization has explicitly prohibited. Existing content filters are static and cannot keep up with the threats your team will face this quarter.

You need a content layer that evaluates prompts and responses against policy you define, redacts sensitive values before they reach a model, and explains its decisions when something is blocked.

What changes when you adopt Touchstone

• Prompt-injection and jailbreak detection

  Catch the attempts that bypass naive filters. Operates at the prompt and response boundary, not after the fact.

• Sensitive-data redaction

  Detect PII and sensitive values, redact before content leaves your environment, and restore for the user when the response returns.

• Organization-defined policy enforcement

  Define prohibited content, tone, and scope rules in your own words. Touchstone evaluates prompts, responses, and tool inputs against them.

• Decision audit

  Every detection, redaction, and block recorded for review and compliance evidence.

How Touchstone operates

1. 01

   Route prompts and responses through Touchstone

   Touchstone runs alongside Gateway. Prompts are evaluated before they leave your environment; responses are evaluated before they reach the user.

2. 02

   Apply detection and redaction

   Prompt injection, jailbreaks, PII, and other categories are detected with policy rules you define.

3. 03

   Enforce or warn

   Block, redact, or warn the user — depending on the policy you configure per category and per tenant.

4. 04

   Audit

   Decisions recorded into the session ledger; exportable to your SIEM.

Component map

• Content classifierMulti-layer classification for prompt injection, PII, and policy categories.
• Redaction engineDetects sensitive values and replaces them with reversible tokens.
• Policy engineOrganization-defined rules evaluated per request.
• Decision auditLogs every classification, redaction, and policy decision.

For the full platform architecture, see /platform.

Who uses it and how

• Compliance / Risk

  Scenario: You must enforce content and privacy controls before content reaches an external model.

  Gain: Redaction at the boundary, audit-grade decision logs, and tenant-scoped policies.

• CISO / Security

  Scenario: Prompt injection and jailbreak attempts are showing up in your logs and you need a real defense.

  Gain: A detection layer that runs at the prompt boundary, with a decision trace you can review.

• Privacy / DPO

  Scenario: You need PII handling that respects tenant isolation and lets you prove what left and what did not.

  Gain: Redaction before egress, restoration for the requesting user, and an immutable record of every action.

See the product surface

Supported systems

Chrysent Touchstone integrates with the providers and frameworks below. Each entry shows the current support status. Documentation for the full integration list is coming soon.

• Gateway (recommended)Generally available
• Custom GuardrailsGenerally available
• Amazon GuardrailsTech Preview
• Azure GuardrailsRoadmap

Where it runs and how data is handled

Chrysent sells self-hosted software only. Software and retained product data remain in customer-owned or customer-controlled infrastructure. Customer-configured external model providers receive provider-bound request data under the customer's provider agreement. This model supports data sovereignty and residency requirements. Data handling and retention are described on the Data handling & deployment page. Security architecture, encryption, and incident response are documented on Security & trust.

Benchmark and evaluation methodology

The first public benchmark study is in preparation. Until it ships, design partners can request the evaluation methodology and a private run report under NDA via the contact form.

Other Chrysent products

• Chrysent AI Gateway

  Private beta

  Chrysent AI Gateway provides centralized control over AI usage across multiple vendors and models.

• Chrysent Bastion

  Private beta

  Chrysent Bastion protects enterprise systems from unsafe AI generated actions.

Common questions

• Is redaction reversible?

  Yes. Touchstone replaces sensitive values with reversible tokens before content leaves your environment, then restores the original values when the response returns to the requesting user.

• Who defines policy?

  You do. Touchstone evaluates prompts and responses against rules your organization defines. Chrysent ships sensible defaults; you can override any of them.

• Does Touchstone add latency?

  Detection and redaction run in parallel with the model call where possible. P50 overhead is reported in the published benchmark study.

• How is sensitive data stored?

  Sensitive values are tokenized in the session ledger; the mapping is encrypted with tenant-scoped keys. See the Data handling page for the full model.