CVE-2026-58449: txtai: unsafe reflection RCE in /reindex endpoint
CVE-2026-58449 affects txtai through 9.10.0, where the /reindex API endpoint resolves a caller-supplied dotted path without an allowlist. Under the stated conditions (API exposed, no TOKEN configured, writable index), the record describes remote code execution as the server process; the record indicates the fix is commit 11b32da, which gates the endpoint behind a new reindex configuration flag.
Overview
CVE-2026-58449 affects txtai through 9.10.0. According to the CNA description, the /reindex API endpoint resolves its function body parameter through txtai.util.Resolver without an allowlist, and the weakness is classified as CWE-94 (Code Injection). The record states the issue is fixed in commit 11b32da, which gates the endpoint behind a new reindex configuration flag. See the VulnCheck advisory and the fix commit.
Impact
- Under the stated deployment conditions, a remote party without authentication could achieve code execution as the server process during reindexing.
- CVSS 3.1 base score 9.8 (Critical) with vector CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H, as provided by the CNA. A CVSS 4.0 score of 9.3 is also listed by the same provider.
- The record does not describe further impacts beyond code execution in the server context.
Affected products and versions
- The record states that txtai through 9.10.0 is affected. The structured affected list in the record is empty, so the range relies on the description text.
- The fix is identified as commit 11b32da. The record does not list a released fixed version, so whether a published release contains the fix is unknown.
- The described risk requires three conditions: the API is exposed, no TOKEN is configured, and the index is configured as writable.
Exploitation status
- The CISA KEV catalog entry is not listed in the record.
- No confirmed exploitation in the wild appears in the sources consulted.
- The EPSS record shows a probability of 0.01354 and a percentile of 0.70876, dated 2026-10-13.
Remediation and mitigation
- Update txtai to a build that includes commit 11b32da, and confirm with the maintainers which release tag contains it.
- Do not expose the txtai API to untrusted networks. Restrict access at the network level and place it behind an authenticating gateway.
- Configure the API token, since the record states that authentication is opt-in and its absence leaves all endpoints unauthenticated.
- Where reindexing through the API is not needed, avoid running with a writable index configuration.
- After upgrading, review the new reindex configuration flag introduced by the fix and keep the capability disabled unless it is required.
Detection guidance
- Review API access logs for requests to the /reindex endpoint from unexpected sources or without expected credentials.
- Monitor the txtai server process for unexpected child processes or outbound activity around reindexing operations.
- Audit deployments for exposed APIs that lack a configured token and that use writable index settings.
Limitations of the available information
- The record has no structured affected or fixed version lists, so the exact released versions carrying the fix are unknown.
- No vendor release notes are included in the record, so the release status of commit 11b32da is unknown.
- The record contains no KEV details, and no public exploit availability is established by the sources consulted.
- The CVSS data come from the CNA provider; no NVD or CISA ADP scores are present in the record.
Frequently asked questions
Is CVE-2026-58449 exploited in the wild?
The record shows it is not listed in CISA KEV, and no confirmed exploitation appears in the sources consulted.
Does the default txtai configuration expose this risk?
The record states exploitation requires an exposed API, no TOKEN configured and a writable index, and that this is not the default configuration.
What should administrators do first?
Restrict network access to the API, configure a token, and update to a build containing commit 11b32da once its release status is confirmed.
This analysis was drafted with AI assistance from verified source data and automatically checked against it.