Description

An arbitrary file upload vulnerability in the /v1/my_drive/batch_upload component of cohere North AI v1.1.5 allows attackers to exeute arbitrary code via uploading a crafted file.

Executive Summary

Technical Details

Mitigations

References

Additional Metadata

{
  "score": "9.8",
  "pubDate": "2026-08-26T19:16:44.927Z",
  "pubdate": "2026-08-26T19:16:44.927Z",
  "executiveSummary": "The cohere North AI v1.1.5 platform contains an arbitrary file upload vulnerability within the /v1/my_drive/batch_upload component. This vulnerability stems from improper validation of user-supplied files, allowing an attacker to bypass security filters and upload malicious content to the server.\nSuccessful exploitation of this flaw enables Remote Code Execution (RCE), granting the attacker the ability to execute arbitrary commands on the underlying host system. This represents a critical risk, as it permits unauthorized system access, data exfiltration, and potential compromise of the entire infrastructure hosting the cohere North AI environment.\nThe vulnerability allows an unauthenticated or authenticated attacker, depending on the implementation context, to achieve full control over the application server by leveraging the lack of file type and content integrity checks. Remediation is necessary to prevent complete system compromise.",
  "technicalDetails": "The vulnerability exists within the /v1/my_drive/batch_upload endpoint of cohere North AI v1.1.5. This component is designed to handle multi-file uploads but fails to enforce sufficient server-side validation regarding file extensions, MIME types, or internal file structure contents.\nThe root cause is an insecure implementation of the file handling mechanism, which assumes the integrity of files uploaded through the batch processing function. By failing to perform comprehensive sanitization or utilize a secure file storage strategy (such as renaming files with non-executable extensions or storing them outside the web root), the application allows the storage of executable scripts or binaries on the filesystem.\nThe attack flow begins with an attacker crafting a malicious payload, such as a web shell or a script designed to interact with the system environment. The attacker then sends a multipart/form-data POST request to the /v1/my_drive/batch_upload endpoint. Because the application does not validate the nature of the file, the server saves the malicious file to a predictable or accessible directory.\nOnce the file is successfully uploaded, the attacker triggers its execution by navigating to the file's path via a standard HTTP request or by inducing the server to process the uploaded file through other system functions. Because the execution occurs in the context of the web server process, the attacker inherits the permissions associated with that user account.\nUpon successful execution, the payload can perform post-exploitation activities, including, but not limited to, internal network scanning, lateral movement, execution of system-level commands, modification of application logic, or complete exfiltration of sensitive data stored within the cohere North AI infrastructure. This vulnerability bypasses standard security controls and requires no further authentication if the endpoint is exposed, or minimal requirements if the endpoint resides behind a broken authentication mechanism.\nThe impact is maximal, as the ability to execute arbitrary code typically leads to a total compromise of the server instance, impacting the confidentiality, integrity, and availability of all data managed by the affected version of the product."
}