Loading vulnerability…
CVE-2026-25879 affects Langroid SQLChatAgent prior to version 0.63.0. The component executed SQL generated by an LLM, and that SQL could be influenced by prompt injection, including indirectly through data returned to the model. Because the generated statement text was executed against the connected database, an attacker able to shape agent input could coerce the agent into issuing unsafe, dialect-specific SQL primitives rather than benign queries. In deployments where the configured database role exposed capabilities such as server-side program execution or filesystem access, this behavior could be escalated from unauthorized SQL execution to command execution on the database host. The issue was addressed in version 0.63.0 by changing SQLChatAgent to default to a SELECT-only sqlglot-parsed statement allowlist and adding a dialect-aware dangerous-pattern blocklist. A configuration option to re-enable the prior unrestricted behavior remains available for explicitly trusted deployments.
Mallory correlates every CVE against your assets, your vendors, and active adversary campaigns. Know which vulnerabilities matter for you, not just which ones are loud.
What it means. What to do now. Patch path, mitigations, and the assume-compromise checklist.
What an attacker gets, and what they’ve been doing with it.
If you can’t patch tonight, do this now.
Patch, then assume compromise.
No public exploits tracked yet. Mallory keeps watching.
No public exploit code observed for this vulnerability.
Products and vendors Mallory has correlated with this vulnerability. Open in Mallory to drill down to specific CPE configurations and version ranges.
Vendor-confirmed product mapping. Mallory continuously reconciles this list against your asset inventory.
11 sources tracked across advisories, community write-ups, and news. New activity surfaces here as Mallory finds it.
A previously fixed Langroid SQLChatAgent prompt-injection-to-SQL-to-RCE vulnerability referenced for comparison because it shares the same defect class and threat model.
A prior Langroid advisory/fix related to blocking `pg_read_file`; it is referenced as the earlier fix that the newly described bypass undermines.
Query your assets running an affected version, and investigate the blast radius.
Every observed campaign linking this CVE to a named adversary.
Malware families riding this exploit, with evidence and IOCs.
YARA, Sigma, Snort, and vendor rules, auto-deployed to your SIEM.
Cross-references every affected SKU, including bundled OEM variants.
Community discussion across Reddit, Mastodon, and other social sources.