Threat Intelligence Database

CVE-2026-55418 is an authorization bypass vulnerability in labring FastGPT, an open source AI knowledge base platform. Prior to version 4.15.0-beta5, two file handlers in FastGPT authorize unrelated resources but then use an S3 object key from the request without verifying team ownership. Since S3 object keys are global within the bucket and tenant IDs appear only as path segments, an attacker can supply another team's key to access their file contents via certain endpoints. This issue is fixed in version 4.15.0-beta5.

CVE-2026-54607 is a Server-Side Request Forgery (SSRF) vulnerability in labring FastGPT prior to version 4.15.0-beta4. The issue arises because the HTTP-tool OpenAPI schema importer only validates the top-level URL before passing it to SwaggerParser.bundle, which fetches remote $ref URLs without internal address validation. This allows an authenticated team member to read internal services or cloud metadata. The vulnerability is fixed in version 4.15.0-beta4.

FastGPT versions prior to 4.15.0 contain an authorization bypass vulnerability in the GET /api/core/ai/record/getRecord endpoint. Authenticated users can access LLM request and response traces belonging to other teams if they know the requestId, due to lack of team scoping. This issue is resolved in version 4.15.0.

FastGPT versions from 4.14.17 up to but not including 4.15.0-beta4 contain a vulnerability that allows an authenticated tenant user to persist a server-owned datasetId from another tenant. This leads to mixed dataset objects and inconsistent ownership checks, enabling cross-tenant read, update, and delete access if object IDs are known. The issue is fixed starting with version 4.15.0-beta4.