We’re thrilled to announce the alpha release of our new CalypsoAI Red Team product. This groundbreaking product allows organizations to launch adversarial attacks against AI models and applications with the click of a button. It combines automated, scalable assessments with over 10,500 static prompts and advanced agentic and operational attacks, including customizable scenarios, to generate actionable vulnerability reports.

CalypsoAI Red Team is available now via an early access program and will be generally available on March 31. For more information, contact our sales team.

We’ve released a powerful new addition to our product: support for OpenAI-compatible APIs. This feature allows you to seamlessly connect and leverage AI models that adhere to the OpenAI API standard, expanding your options beyond OpenAI's offerings.

Key benefits:

We’ve added more filtering options to our Prompt History logs UI to make it easier to find exactly what you’re looking for. You can now:

We’ve made two major security enhancements to our product: Multi-Factor Authentication (MFA) and SAML 2.0 support. These additions significantly boost your account security and streamline enterprise-level access management.

MFA adds an extra layer of security to your account by requiring a second form of verification in addition to your password. To enable MFA in your SaaS or on-premises deployment:

Super admins can also reset MFA for individual users who’ve been locked out or lost their token.

SAML 2.0 integration allows enterprise customers to use their existing identity provider for single sign-on (SSO), simplifying user management and enhancing security. Benefits:

To get SAML 2.0 set up for your organization, please contact support. You’ll need to provide:

Blank user column for scans. The user column for scans was previously blank but now it will show the name of the API token used

Chat should still be disabled after clicking New Chat. If no providers are configured, we show a banner in the chat view stating this and prevent the user from sending a prompt until a provider is chosen. Previously, the user could click New Chat to get around this and end up in a broken state. Now, clicking New Chat will still display the banner, but disable the chat until a provider is chosen.

Show error log for custom providers. Our new error log display in Model Configurations was only showing for our set providers. Now it will also show if there is an error when creating a custom provider.

Improvements to Keycloak MFA.Users that previously had MFA enabled will have it disabled if the setting is turned off across the organization.

Expired badge still shows after generating new invite link. Previously, when an “Invite” link was expired, clicking Generate Invite Link would generate a new valid link, but the “Expired” badge would still be shown in the table. Now it reverts back to the “Invite” badge.

Broken page when navigating to the groups tab of Model Configurations. If no groups were set up and you navigated to the Groups tab of the Model Configuration screen, the page would enter an error state. This has been fixed.

Record Prompts. Toggling off the Prompt Logs switch on the Logs screen will now stop the recording of prompts across the organization.

We’ve sunsetted the “Disclosures” feature from Chat settings, due to lack of adoption