August 4, 2026 WitnessAI Release

Note: New and updated features may not be immediately available on your dedicated deployment due to participation in limited availability releases, beta programs, or regional configurations.

If you don't see a feature you expect, please reach out to Customer Success. We're happy to help!

New Features

Amazon Bedrock & Amazon Q

Witness now monitors Amazon Bedrock and Amazon Q interactions with full prompt, response, and agentic observability. Supports API key authentication for Bedrock access. All guardrails, DLP policies, and behavioral analytics apply to these interactions.

Microsoft Copilot for Cowork

Witness now monitors Microsoft Copilot used within the Cowork integration, capturing prompts and tool-call interactions for compliance and observability.

Lists Management Landing Page

A new Lists Management landing page provides a centralized view of all configured lists, improving navigation and management of allow/deny/watch lists.

Conversation Export (JSON / CSV)

Users can now export conversations in JSON or CSV format directly from the Conversations page, enabling downstream analytics and compliance record-keeping.

Feature Updates

Home Dashboard

The redesigned Home Dashboard displays new at-a-glance views of AI activity with tabs. You can quickly view details of Employees using AI, Agents performing tool actions, the Applications you protect, and Value delivered with WitnessAI without having to navigate and filter in each separate dashboard.

ChatGPT Pro / High Reasoning Mode — Full Policy Support

Full policy enforcement support for ChatGPT Pro and High (extended thinking) modes, including routing, blocking, warning injection, and response redaction. Witness also distinguishes between Personal and Work ChatGPT accounts for appropriate policy application.

Claude Code — MCP Tool Resolution & Agent Tool Use Policy

Witness now resolves MCP tool names in Claude Code agentic sessions and enforces agent tool use policies on traffic. This enables compliance teams to control which MCP tools agents may invoke.

Global App Allowlist — Empty-List Warning & API

When the Global App Allowlist is enabled with an empty list, users now see a clear warning that all applications will be blocked. Backend API support for managing the allowlist has also been added.

AI Value Dashboard — Configurable Token Cost Rates

Administrators can now configure custom token cost rates in the AI Value Dashboard, enabling accurate cost tracking aligned to enterprise contract pricing.

Salesforce AI Bots

Salesforce AI Bot traffic is now fully monitored and visible in the Witness platform.

Kiro CLI Proxy Routing

Kiro CLI traffic is now routed through the Witness proxy, enabling full observability and policy enforcement for Kiro-based AI interactions.

JetBrains AI, Codex, and Claude Code Subprocess Traffic Capture

Subprocess traffic from JetBrains AI Assistant, Codex CLI, and Claude Code spawned processes is now captured and associated with the originating session.

Writer.com Built-in Agent Library Observability

Witness now provides observability into Writer.com's built-in agent library interactions, extending coverage to enterprise writing automation workflows.

Witness Anywhere Registration Improvements

Multiple improvements to the Witness Anywhere registration process:

Performance, Reliability, & API Updates

Policy evaluation latency at p95 was elevated, causing occasional slowdowns at the proxy layer. This release reduces high-end latency in the policy enforcement pipeline.
The Discovery overview page was taking over 10 seconds to load for some large tenants. Response times are now significantly faster.

Fixed Issues

Registration failure when the Windows username contains spaces
Jamf Marker Profile now uses email as the user identifier for consistent registration
Netskope Destination Profiles are now an opt-in configuration rather than always-on
Home Dashboard v2 — Tab Consistency (Generally Available)

Tab navigation in the Home Dashboard v2 now behaves consistently across all views, resolving visual and state inconsistencies.

Policy Management

Saving a policy could inadvertently create a duplicate policy.
Agent tool use policy lists could display incorrectly.

Conversations

Prompts might not display properly in the conversation context view for very long threads.
The conversation details panel was closing when clicking away from a dropdown menu inside it. The panel now stays open when interacting with menus within it.

Security

Resolved a server-side validation gap that could allow inconsistent user records. User creation now enforces proper input validation.

Known Issues

App Catalog: Download CSV file has Policy column with incorrect name. [Object][Object]

VS Code: Customized Warn message not appearing in inline functionality.

MS Excel Desktop: Not able to access ChatGPT plugin on MS Excel thick client.

MS Office web apps: Not able to create blank documents without using copilot.

Google AI Mode: If you have Google AI Mode (www.google.com) in your App Catalog, but don't add it to the allowlist, other apps dependent on Google AI Mode might not work (like zoom). This works as intended, but can surprise customers.

Breaking Changes

No known breaking changes in this release.

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