July 31st, 2026

AI Discovery (Workforce)

AI Workforce Device tracking — Devices used by your workforce for AI interactions are now tracked with proper user linking, giving you better visibility into which devices are accessing AI tools.
Browser extension monitoring improvements — The browser extension now monitors embedded content and iframes, automatically bypasses internal network addresses to avoid interfering with your tools, and correctly identifies users even without Chrome Sync enabled.

Amazon Bedrock response decoding — The endpoint agent now fully decodes Amazon Bedrock API responses across all supported formats, enabling accurate logging and guardrail evaluation for Bedrock-based AI workloads.

Token estimation for workforce logs — AI token usage is now estimated for workforce logs, giving you better insight into consumption patterns even when exact token counts are not provided by the AI service.
Guardrail evaluation and request metrics in workforce logs — Workforce employee log details now show guardrail evaluation results alongside request metrics including token counts, sentiment, and subjectivity scores.
Reduced log noise from repeated website visits — Repeated visits to the same AI website no longer generate redundant log entries, keeping your activity logs cleaner and more meaningful.
Fixed workforce data integrity during rescans — Fixed an issue where workforce consumption data could be incorrectly deleted during narrow rescan windows for certain integrations.

Guardrail block page in browser — When a guardrail blocks content, users now see a dedicated block page in their browser with clear explanations of which guardrail was triggered and why.
Real-time policy updates via WebSocket — The endpoint agent now receives policy and rules updates in real time, so changes you make to governance policies take effect immediately without waiting for the next polling interval.
Improved guardrail evaluation resilience — Guardrail evaluation no longer fails entirely if a topic guardrail is missing, undeployed, or still populating. Unavailable guardrails are skipped and reported in a clear list while the rest of your evaluation proceeds normally.
Smarter policy creation workflows — Policy creation now pre-selects scope and target based on your current context (employee, group, device, or project), making it faster to create policies without manual selection.
High-volume resource policy notifications — Policies that match a high volume of resources now show aggregate match counts in notification history, giving you a clearer picture of policy impact.
Improved policy check history accuracy — Policy check history now records entries even when no matches are found, and debounces empty results to at most one per day to reduce noise while maintaining visibility.

OWASP AI Security Verification Standard (AISVS) support — OWASP AISVS is now available as a security framework you can select and filter by in findings and reporting, helping you assess your AI security posture against this industry standard.
Findings check summary statistics — A new findings check summary section shows total checks, passed, failed, and failure rate statistics at a glance for faster posture assessment.
Improved framework mapping in findings — Findings now display framework mappings more clearly with better logo matching and proper handling of empty states.

Refreshed AI Dashboards — AI Dashboards for Workforce and Workload have been reorganized into clearer sections — Risk Trends, Risk Summary, Token Usage, Log Volume/Activity, and Usage Breakdown — with improved chart tooltips, platform logos, and slide-out drawers that keep you in context when exploring data.
Standardized sorting across resource lists — All resource lists including Models, Agents, Prompts, Policies, Topics, Guardrails, Employees, Devices, Applications, and Logs now have consistent sorting options, defaulting to Risk Score where applicable.
International character support — You can now use accented Latin characters, Japanese scripts (Hiragana, Katakana, Kanji), and additional Unicode characters in names, descriptions, and other input fields across the platform.
Fixed Microsoft 365 OAuth2 authentication — Fixed an issue where the Microsoft 365 popup authentication flow could fail in certain browser security configurations, preventing you from connecting your integration.
Improved disabled UI accessibility — Disabled buttons, inputs, and tabs are now more visually distinct and legible across the application.
Default project auto-creation — New organizations now automatically have a default project created, so you can start using the platform immediately without manual setup.