Popular npm Library Used in AI and NLP Projects Exposes Systems to RCE

A critical remote code execution vulnerability has been found in the popular JavaScript library expr-eval, affecting thousands of AI and NLP projects that rely on it for mathematical expression evaluation and natural language processing. The flaw, tracked as CVE-2025-12735, arises from a design issue in the Parser class’s evaluate() method, enabling an attacker to define arbitrary functions within the parser’s context and execute system-level commands from user-controlled input.

According to the SSVC framework, the Technical Impact is Total, signalling that adversaries could gain total control over the software’s behaviour. The vulnerability was disclosed on 7 November 2025, with updates arriving on 9 November 2025, and is being coordinated by UKO in collaboration with GitHub Security and npm, as disclosed by CERT/CC Note.

Remediation paths include applying the security patch from Pull Request 288 in the expr-eval repository or upgrading to the latest patched version, with expr-eval-fork v3.0.0 addressing this flaw; organisations are urged to audit dependencies and patch promptly to protect production AI and NLP systems. according to CERT/CC Note