Zero-Authorization API Flaw Exposed Schemata’s DoD Training Data Across Tenants

A serious authorization flaw in Schemata’s API exposed sensitive military training data, user records, and direct file links across customer tenants. The issue affected Schemata, an AI virtual training company that works with defense and enterprise customers.

The bug allowed a low-privilege account to access data outside its own organization. In practical terms, the platform failed to check whether a user belonged to the tenant whose data they were requesting.

Strix, an open-source AI security testing tool, found the issue while testing Schemata’s application. The researchers said the flaw exposed user lists, enrollment information, military base details, organization data, course metadata, and links to confidential training material.

What happened

Schemata builds AI-driven virtual training and simulation tools for defense and enterprise use. Its platform creates immersive 3D learning environments, AI-guided instruction, and training analytics for high-stakes work.

That made the API flaw especially sensitive. The vulnerable endpoints were tied to a platform that handled military training courses, service member data, and proprietary training content.

Strix said its agent started with a normal low-privilege account. It mapped reachable API endpoints, replayed requests, and found that the backend returned platform-wide data instead of limiting responses to the test account’s own tenant.

Detail	What researchers reported
Company affected	Schemata
Platform type	AI-powered virtual training and simulation
Main issue	Missing tenant isolation and API authorization checks
Access needed	Low-privilege authenticated account
Data exposed	User records, enrollment data, base information, course data, and S3 file links
Patch status	Researchers said the issue was fixed before publication

Why the flaw was so dangerous

This was a multi-tenant authorization failure. Multi-tenant platforms separate data between customers, organizations, agencies, or teams. When those checks fail, one user can sometimes view data belonging to everyone else.

According to Strix, the Schemata API returned global data from several endpoints. A user-listing endpoint exposed names, email addresses, enrollments, and the military bases where service members were stationed.

That information could support targeted phishing, identity profiling, harassment, or doxing. Military personnel data carries extra risk because attackers can use it to identify units, roles, locations, and relationships between people and organizations.

Training content was also exposed

The exposure did not stop at personal information. Researchers said course and organization endpoints leaked metadata and direct AWS S3 links to training documents.

The exposed materials reportedly included hundreds of military training files. Some related to naval maintenance training, while others involved Army field manuals and explosive ordnance procedures.

Even when documents do not qualify as classified, this type of material can still create operational security concerns. Training content can reveal procedures, equipment handling, maintenance steps, or internal readiness workflows.

• User names and email addresses
• Enrollment and training participation records
• Military base information
• Organization and course metadata
• Direct AWS S3 links to training material
• Proprietary 3D training content
• Military manuals and procedural documents

Write access raised the risk further

Strix also reported that some write-enabled routes lacked proper authorization checks. That means an attacker may not have been limited to reading sensitive data.

If abused, weak write controls could allow a malicious user to modify course content, alter training records, or delete material. In a defense training environment, that risk goes beyond privacy.

Training systems need strong integrity controls because personnel may rely on them to learn procedures, practice maintenance, and prepare for field tasks. Tampered content can mislead users or damage trust in the training platform.

The disclosure took months

Strix said it first reported the issue to Schemata on December 2, 2025. The researchers described the flaw as critical and sent follow-up messages after the initial disclosure.

The issue remained unresolved for months, according to the disclosure timeline. Schemata later acknowledged the exposed endpoints and patched the bug on May 1, 2026, after a final notice before publication.

The researchers said they verified the remediation. That means the specific cross-tenant access path they reported no longer worked after the fix.

Date	Event
December 2, 2025	Strix privately reported the vulnerability
Following months	Researchers sent follow-up warnings about the issue
May 1, 2026	Schemata acknowledged the exposed endpoints and patched the flaw
After patch	Strix said it verified remediation

Why this matters for defense contractors

Schemata’s own website describes its platform as a virtual training system for defense operators and enterprise teams. The company also highlights secure field operations, AI guidance, instructor analytics, and immersive 3D training.

Owl Ventures said in 2025 that Schemata had been awarded $3.4 million in Department of Defense contracts over the previous eight months. The same report said the company raised $5 million in seed funding.

Defense contractors face strict cybersecurity expectations when they handle federal contract information or controlled unclassified information. CMMC and NIST SP 800-171 both focus on protecting sensitive government-related data in nonfederal systems.

What affected customers should ask

The key question now is whether anyone accessed exposed data before the fix. A patch closes the vulnerability, but it does not answer what happened during the exposure window.

Customers should ask for access logs, tenant access reviews, S3 access history, and any evidence of unusual API activity. They should also ask whether affected users will receive formal notification.

For platforms that serve government or defense customers, authorization needs testing at every route. A login system does not protect data if the backend fails to check which tenant a user belongs to.

• How long were the vulnerable endpoints exposed?
• Did any unknown accounts access cross-tenant data?
• Were S3 file links accessed or downloaded?
• Were write-enabled routes abused?
• Were affected users and organizations notified?
• Were audit logs preserved for review?
• Has Schemata completed a third-party security review after the patch?

FAQ