Attack Chain

1. An attacker prepares a malicious input schema designed for datamodel-code-generator.
2. Within this schema, the attacker embeds a specially crafted customBasePath value.
3. The customBasePath value includes embedded newline characters and a dot-free Python expression (e.g., os.system('malicious_command')).
4. A legitimate user or automated system processes this malicious input schema using datamodel-code-generator prior to version 0.70.0.
5. During the code generation process, the datamodel-code-generator tool insecurely embeds the malicious customBasePath verbatim into a from ... import ... statement of the output Python module without validating the identifier.
6. When an application subsequently imports this newly generated, compromised Python module, the embedded malicious Python expression is executed.
7. This execution leads to remote code execution (RCE) on the system where the generated module is imported, allowing the attacker to run arbitrary commands.

Impact

Successful exploitation of CVE-2026-63720 grants an attacker remote code execution capabilities on the system running the vulnerable datamodel-code-generator instance or any application that imports code generated from a malicious schema. This allows attackers to compromise the integrity and confidentiality of data, establish persistence, and potentially pivot to other systems within the environment. The severity of the impact depends on the privileges of the affected system, but generally represents a critical breach. No specific observed victims or targeted sectors are mentioned in the advisory, but any organization using affected versions of datamodel-code-generator to process untrusted schemas is at risk.

Recommendation

• Patch datamodel-code-generator to version 0.70.0 or later immediately to remediate CVE-2026-63720.