/

CVE-2025-61165

Severity:

CRITICAL RISK

(9.8)

Arbitrary File Upload (CWE-434) in Cohere North AI

CVE-2025-61165

Deferred

Vulnerability type

Category:

Unrestricted File Upload

CWE:

CWE-434

Impact:

Remote Code Execution

Severity:

CRITICAL RISK

(9.8)

CVSS version:

3.1

Catch this CVE while code is being written

Detect vulnerable code directly in pull requests, so CVEs are fixed before merge.

Summary

A critical arbitrary file upload vulnerability in Cohere North AI v1.1.5 affects the /v1/my_drive/batch_upload component. If an attacker can reach this upload functionality and submit a crafted file, the application may execute attacker-controlled code.

Affected context

The vulnerability occurs in deployments that expose or otherwise permit access to the /v1/my_drive/batch_upload endpoint and do not sufficiently restrict, validate, store, or process uploaded files. Risk depends on the endpoint's network exposure, authentication controls, file handling behavior, and the privileges of the component processing the upload.

Exploit preconditions

  • Attack surface: The /v1/my_drive/batch_upload file upload component
  • Required access: Access sufficient to submit files to the upload endpoint;

authentication requirements depend on deployment configuration

  • Trigger point: Processing or handling of a crafted uploaded file
  • Why exploitation is possible: The upload path may accept attacker-controlled

file content without adequately restricting file types or preventing execution

Why this happens

The application does not adequately validate and constrain uploaded files before storing or processing them, allowing attacker-controlled content to cross the trust boundary and be treated as executable code. This maps to CWE-434: Unrestricted Upload of File with Dangerous Type.

Potential impact

  • An attacker may execute arbitrary code in the application environment.
  • The application or its underlying host could be compromised.
  • Sensitive files, credentials, or stored user data may be accessed.
  • The vulnerability could enable persistence, lateral movement, or service disruption.
  • Data could be modified, deleted, or exfiltrated.

General mitigation guidance

Mitigation generally focuses on treating every uploaded file as untrusted data

and preventing uploaded content from becoming executable.

Common defensive patterns include:

  • Allowing only narrowly defined file types based on validated content rather than

filename or client-provided MIME type.

  • Applying size, quantity, filename, and archive-expansion limits.
  • Storing uploads outside executable or served directories with restrictive

permissions.

  • Using isolated processing environments with minimal filesystem, network, and

operating-system privileges.

  • Generating controlled server-side filenames and separating uploaded content

from application code.

  • Scanning and validating files before downstream processing.
  • Requiring authentication and enforcing authorization for upload operations.
  • Applying secure-by-default web server and runtime policies that disable execution

of uploaded content.

How teams detect issues like this

Detection approaches commonly include:

  • Static analysis for upload handlers that trust filename extensions, MIME types,

or client-supplied metadata.

  • Code review checks for uploads stored beneath executable or web-served paths.
  • Searching for direct use of uploaded filenames in filesystem paths or shell

commands without canonicalization and validation.

  • Monitoring upload volume, rejected file types, unusual extensions, and requests

to the batch upload endpoint.

  • Alerting on process creation, unexpected interpreter use, or outbound connections

originating from the upload-processing component.

  • Enforcing CI and deployment policies that require isolated upload storage and

least-privilege execution.

Automated security review tools may help identify insecure upload handling patterns,

but findings should be validated against the deployment architecture.

Sources

SourceLink
cohere.comcohere.com
north.comnorth.com
github.comgithub.com/bdadoa/Cohere---North-AI-1.1.5---Vulnerabilities/blob/main/CVE-2025-61165.md

Frequently asked

questions

It may be remotely exploitable when the batch upload endpoint is network-accessible

and an attacker can submit files through it. The practical scope depends on

authentication, authorization, and network exposure.

Not necessarily. It depends on how the endpoint is deployed and whether access

controls require a valid user or service identity.

Treat uploads as untrusted data, allow only required file formats, validate content,

store files outside executable paths, and process them in a least-privileged,

isolated environment.

Combine strict upload validation with authorization, non-executable storage,

resource limits, monitoring, and CI policies that flag unsafe file-handling

patterns.

Find out what's already
exploitable in your codebase.