Cohere Python
Monthly
Arbitrary file write in the Cohere Python SDK 5.11.0 through 7.2.0 arises from unvalidated tar archive extraction in the _s3_models_dir_to_tarfile helper used by the SDK's cohere_aws SageMaker integration. An attacker who can place a crafted model archive into the S3 prefix the SDK reads from can embed absolute paths or ../ traversal members that, once tarfile.extractall runs on the victim host, overwrite files outside the intended model directory. Exploitation is precondition-heavy: it requires the affected SageMaker model-archive flow to be in use, attacker write access to the victim's S3 model location, and the victim to trigger the extraction (passive user interaction); there is no public exploit identified at time of analysis and no confirmed actively exploited (CISA KEV) status. EPSS and KEV data were not provided, and the impact is scored as integrity-only (no confidentiality or availability effect), making this a medium-severity, precondition-dependent flaw.
Arbitrary file write in the Cohere Python SDK 5.11.0 through 7.2.0 arises from unvalidated tar archive extraction in the _s3_models_dir_to_tarfile helper used by the SDK's cohere_aws SageMaker integration. An attacker who can place a crafted model archive into the S3 prefix the SDK reads from can embed absolute paths or ../ traversal members that, once tarfile.extractall runs on the victim host, overwrite files outside the intended model directory. Exploitation is precondition-heavy: it requires the affected SageMaker model-archive flow to be in use, attacker write access to the victim's S3 model location, and the victim to trigger the extraction (passive user interaction); there is no public exploit identified at time of analysis and no confirmed actively exploited (CISA KEV) status. EPSS and KEV data were not provided, and the impact is scored as integrity-only (no confidentiality or availability effect), making this a medium-severity, precondition-dependent flaw.