Sign in
← AI PulseSep 16, 2026

Wire · news · Multi-source brief

OpenAI Disrupts Russian-Speaking Malware Development Operation "ScopeCreep"

OpenAI banned a cluster of ChatGPT accounts operated by a Russian-speaking threat actor who used models to develop and refine Windows malware, debug code, and set up command-and-control infrastructure.

By Illumora Editorial

Source · Sep 16, 2026, 3:56 PM · On Illumora · Sep 16, 2026, 4:13 PM

Media from the primary source — shown here so you can stay on Illumora.

Synthesized from multiple allowlisted primaries on the same event. Lanes →

Brief drafted by Illumora’s editorial model from the linked primary source. Ops desk reviews flagged pieces. How we write →

Read the source →OpenAI Safety — Operation “ScopeCreep”: Russian-speaking malware development
Save

OpenAI identified and banned a cluster of ChatGPT accounts in June 2025 that were associated with a Russian-speaking threat actor. This actor, dubbed "ScopeCreep," utilized OpenAI's models to assist in developing and refining Windows malware, debugging code across multiple languages, and configuring command-and-control infrastructure.

The operation focused on a trojanized "crosshair" gaming tool, and the actor demonstrated knowledge of Windows internals and operational security behaviors.

Key Points

  • OpenAI banned ChatGPT accounts in June 2025 that were linked to a Russian-speaking threat actor.
  • The actor used OpenAI models to develop and refine Windows malware, debug code, and set up command-and-control infrastructure.
  • The operation, named "ScopeCreep," involved distributing malware through a publicly available code repository impersonating a legitimate crosshair overlay tool.
  • The threat actor employed operational security tactics, including using temporary email addresses for ChatGPT accounts and limiting each account to one conversation for incremental code improvements.
  • Malware capabilities included privilege escalation, persistence, credential harvesting, and remote access, with features like C2 payloads designed to avoid signature-based detections and stealthy execution via DLL side-loading.
  • OpenAI detected the activity through its scaled cyber abuse detection process and coordinated with the code hosting provider to take down the malicious repository.
  • The threat actor used the model to iteratively develop malware features, such as debugging Go code for HTTPS requests and developing PowerShell commands in Go to modify Windows Defender settings.

Context

According to OpenAI, the threat actor's approach to operational security involved using temporary email addresses to sign up for ChatGPT accounts. Each account was used for a single conversation about an incremental code improvement before being abandoned for a new one. The ScopeCreep malware was distributed via a public code repository that impersonated a legitimate gaming tool, Crosshair-X, leading unsuspecting users to download and execute malicious files.

Why It Matters

This incident highlights how threat actors may attempt to leverage AI models to accelerate malware development, while also demonstrating the potential for AI providers to detect and disrupt such activities through robust abuse detection processes. Builders and security practitioners can note the specific techniques used by the threat actor and the detection methods employed by OpenAI.

What To Do

  • Note the operational security tactics used by the threat actor, such as temporary email addresses and single-use accounts.
  • Review the examples of model interactions, such as debugging Go code and developing PowerShell commands, to understand how models can be misused for malicious development.
  • Consider the detection methods mentioned, such as scaled cyber abuse detection processes, as a reference for identifying similar activities.
  • Watch for updates from OpenAI regarding their ongoing efforts to disrupt malicious uses of AI.