CVE-2026-85689: Improper Neutralization of Special Elements used in an SQL Command ('SQL Injection') in llmware-ai llmware

CVE-2026-85689 describes an SQL injection vulnerability in llmware 0.4.6 within the collection-database layer (llmware/resources.py). The vulnerability is due to direct string interpolation of filter and lookup values into SQL WHERE clauses without parameterization or escaping, despite an allow-list check on keys. The filter validator does not sanitize values, allowing attacker-controlled input via public API methods such as Library.block_lookup and Query.text_query_with_custom_filter or text_query_by_author_or_speaker. This leads to unauthorized data disclosure across documents or collections. On PostgreSQL, the vulnerability permits boolean- and UNION-based SQL injection techniques.

An attacker with low privileges can exploit this vulnerability remotely without user interaction to bypass intended data filters and access unauthorized rows in the database. This results in cross-document and cross-collection data disclosure. On PostgreSQL, the vulnerability further allows advanced SQL injection techniques that could lead to broader data exposure or manipulation.

Patch status is not yet confirmed — check the vendor advisory for current remediation guidance. Until an official fix is available, avoid using affected versions or restrict access to the vulnerable API methods. Implement input validation and parameterized queries to prevent SQL injection.