September 22, 2026: WitnessAI Release

Replit Support for Prompt, Response, Intention, and Tool Call Observability is included in this release.

Jitera AI is now supported for prompt, intention, and tool call observability. Jitera activity now appears in Conversations. This release delivers observability, policy enforcement is not yet available.

Gemini Enterprise and Vertex AI Search are now supported for prompt, intention, and tool call observability. This release delivers observability only, policy enforcement is not yet available.

The Spend Dashboard displays AI token usage and cost, with attribution according to seat-based subscriptions, metered token usage, and on-demand usage exceeding plan quotas. Customizable settings include subscription providers, seats, seat pricing, and included usage. Settable default rates for unknown pricing enables reporting without price gaps.

Setting configuration is restricted to administrative roles.

Claude account type recognition has been updated. Enterprise, Team, and Personal account types will be displayed for all Claude activity.

This makes it easier to separate sanctioned enterprise Claude use from personal accounts when reviewing usage and building policy.

VIP user handling is clearer throughout the console. Alerts for VIP users and their alert detail remain visible, but attempts to drill into a VIP user's conversations now explain that conversations from VIP users cannot be viewed, rather than failing silently or returning an empty result.

Validation results are easier to work with. The re-run control is now labeled "Re-validate" to match the "Run Validation" step, a redundant Reset control has been removed, and completed validation runs can be downloaded as CSV or JSON. Adding a template to an existing guardrail now appends its rules to the rules already there, with duplicate rules skipped and a confirmation prompt naming the target guardrail, instead of replacing them.

No performance, reliability, or API updates in this release.

Fixed an issue where the Applications and Departments filters on the Home dashboard could display raw internal application and department values instead of their friendly labels.

Fixed an issue where three-digit percentage values could fail to save in the Value dashboard configuration.

Fixed an issue where the Shadow AI prevention statistic on the AI Value dashboard may not have been pro-rated to the selected timeframe, which could overstate the figure for shorter date ranges.

Fixed an issue where the documentation search control could be missing from the Home dashboard header.

Fixed an issue where tenants with no custom data types could see a "Failed to fetch data types" message on the Alerts page, with the Data Type filter unavailable and built-in data types missing from the dropdown.

No known issues in this release.

No breaking changes in this release.

Please contact Customer Success if you have questions or need support with this release.