CVE-2026-41496: PraisonAI SQL Injection Impacts Multiple Backends

The National Vulnerability Database has detailed CVE-2026-41496, a high-severity SQL injection vulnerability affecting PraisonAI’s multi-agent teams system. This issue, present in versions prior to PraisonAI 4.6.9 and PraisonAI Agents 1.6.9, stems from a critical oversight in input validation. While a previous fix for CVE-2026-40315 addressed validation for SQLiteConversationStore, nine other database backends were left exposed.

These vulnerable backends—including MySQL, PostgreSQL, Turso, SingleStore, Supabase, and SurrealDB, along with their async counterparts—directly embed table_prefix into f-string SQL queries without proper sanitization. The National Vulnerability Database reports 52 distinct unvalidated injection points across the codebase sharing this root cause. Furthermore, the postgres.py module exacerbates the risk by accepting an unvalidated schema parameter directly used in DDL operations, opening the door for even more severe database manipulation. This allows for arbitrary SQL execution, leading to potential data exfiltration, modification, or even complete database compromise.

This vulnerability, with a CVSS score of 8.1 (High), highlights a fundamental security flaw (CWE-89) that could allow authenticated attackers to gain significant control over the underlying database. Organizations leveraging PraisonAI with any of the affected database backends are at severe risk. The issue has been patched in PraisonAI version 4.6.9 and PraisonAI Agents version 1.6.9.