SAFi v1.5.0: Enterprise Features for Governed AI

Hi everyone, SAFi v1.5.0 is here and is packed with enterprise features.

This release strengthens the controls organizations need to deploy AI responsibly, including sensitive-data protection, enterprise identity management, security hardening, team collaboration, human approval workflows, and improved administrative tooling.

Here is what is included:

Sensitive-data protection

SAFi now screens prompts locally for sensitive information before they are sent to external model providers. Detection covers Social Security numbers, credit card numbers, bank routing numbers, and related identifiers.

When SAFi blocks a prompt, it gives the user a clear explanation instead of returning a generic error. Legal holds also suspend applicable record-destruction and redaction routines.

Enterprise identity and access management

Organizations can connect SAFi to their existing identity infrastructure through:

1. Microsoft Entra ID single sign-on
2. TOTP multi-factor authentication
3. SCIM 2.0 user and role provisioning
4. Automated account deprovisioning
5. DNS-based domain verification
6. Tenant-aware workspace routing

These controls help administrators manage access throughout the user lifecycle.

Security and integrity improvements

Administrators can use Verify this Install to compare a deployment with SAFi’s signed release manifest.

The release also adds:

1. Screening of external tool results for prompt-injection patterns
2. SHA-256 package and dependency verification
3. Updated server and document-parsing dependencies
4. Unprivileged Docker container execution
5. Credential removal from subprocess environments
6. Removal of dynamic runtime npm package installation

Together, these changes reduce deployment risk and strengthen the boundary between models, tools, credentials, and infrastructure.

Team collaboration

Users can share conversations, research threads, and knowledge folders with colleagues or designated groups.

Granular permissions control who can view or edit shared resources, while a dedicated Shared with Me view keeps collaborative material separate from personal conversation history.

Attention Inbox and approval workflows

The new Attention Inbox provides one place to review:

1. Account and role access requests
2. Tool permission requests
3. Security notifications
4. Compliance alerts

Access to privileged tools can require approval from a second authorized administrator, supporting separation of duties. Requesters can also track the status of their submissions.

Expanded model catalog

SAFi v1.5.0 introduces additional model integrations and stable, versionless model identifiers. This allows backend model versions to change without breaking saved prompts, bookmarks, or configurations.

Model names and any Zero Data Retention or HIPAA-related routing claims should be confirmed against the production provider catalog and deployment configuration before publication.

Document exports

Governed answers and completed conversation turns can be exported in several standard formats:

1. PDF
2. Microsoft Word
3. Microsoft Excel
4. Markdown

This makes it easier to move reviewed outputs into reports, records, and operational workflows.

Scheduled email digests

Agents can run on daily or weekly schedules and deliver summaries by email. Digest content is converted from Markdown into branded HTML for a cleaner reading experience.

Progressive Web App support

SAFi can be installed through supported modern browsers on desktop and mobile devices, providing a more application-like experience without requiring a separate app-store installation.

The service worker is intentionally limited to static assets. Governance state, policy definitions, and ledger records are not cached.

Updated platform support

SAFi v1.5.0 also includes the following administrative updates:

1. MySQL 8.4 LTS as the standard supported database target
2. Python 3.11, 3.12, and 3.13 support
3. HTTPS required for SCIM 2.0 endpoints
4. Vetted Python modules and standard MCP servers for tool integrations

SAFi v1.5.0 is a significant step toward giving organizations practical control over how AI systems access data, use tools, collaborate with people, and produce auditable results.

Review the release notes, test the upgrade in your environment, and share your feedback with the SAFi community.