Prior to release 1.2.14, the UbuntuCorpusTrainer in ChatterBot extracts archives into a predictable directory located at the user’s home. The extraction process uses a check‑then‑create pattern followed by a tar.extractall call, which follows symbolic links. A local attacker who can pre‑plant a symlink at the expected path can cause the archive to write files through that symlink into an arbitrary location of the attacker’s choosing, effectively allowing arbitrary file modification or overwriting of critical files. The weakness is classified as CWE-367 and CWE-59, involving race conditions and symlink exploitation.

The vulnerability affects the ChatterBot machine learning conversational engine produced by gunthercox:ChatterBot. All releases prior to version 1.2.14 are vulnerable. The issue is fixed in release 1.2.14 and later, so any deployment using an earlier version must be upgraded.

The CVSS score of 5.5 indicates a moderate severity; the EPSS score is < 1%, indicating a low exploitation probability, and the vulnerability has not been listed in the CISA KEV catalogue. Because the flaw requires local file‑system access and the ability to supply an archive to the trainer, the attack vector is inferred to be local. An attacker would need to execute or influence the ChatterBot process or place files in the user’s home directory, limiting the threat to environments where untrusted code runs with the same user access level or where an attacker has local device access.