Skip to main content

March 2, 2026

Highlights​

This release introduces Message Sanitization for protecting sensitive information and significantly improves voice dictation reliability.

New Features​

Message Sanitization​

A comprehensive system for detecting and protecting sensitive information in chat messages before they reach the AI model. Organization admins can configure policies to allow, redact, or block messages containing personal data, credentials, and other confidential information. To enable this for your organization, contact the Devs.ai team.

  • Detection rules per entity type with configurable confidence thresholds
  • Preset policies: Strict, Balanced, and Internal-Friendly
  • Built-in test console to validate policies before saving
  • Allow list for safe terms that should bypass detection
  • Audit logging for all sanitization actions
  • Support for 50+ entity types across multiple regions

Agent Avatar​

  • New consistent agent avatar design for a unified look across all chat interfaces
  • Support for custom images on agent avatars

Improvements​

Voice Dictation​

Multiple improvements to the voice dictation experience:

  • Dictation no longer auto-submits when you pause speaking
  • Your edits are preserved when dictation resumes
  • Proper cleanup when leaving a conversation

Chat UI​

  • Improved styling on pinned chat items

Skills​

  • File import support for skills with updated content limits

Browser Tools​

  • Better reliability and error handling when connecting to browser-based tools

Bug Fixes​

Various fixes to authentication flows, settings persistence, and chat reliability.

⌘J