Ledger / Feb 2025
DPRK-linked accounts use ChatGPT for intrusion-tool research and RDP brute-force tooling
Confirmed
OpenAI, the model provider, states the accounts used its models for this work.
- Model
- ChatGPT / GPTOpenAI · closed
- Agent harness
- Nonechat or tool use only
- Attributed country
- North Koreaas attributed by OpenAI
- Sectors hit
- crypto, governmentUnspecified; interest in cryptocurrency and cyber intrusion tools
Accounts OpenAI links to Kimsuky (VELVET CHOLLIMA) and BlueNoroff (STARDUST CHOLLIMA) used ChatGPT for coding help, research on intrusion tools, RDP brute-force tooling and RAT development. The date is the report month.
“The banned accounts primarily used our tools to pursue information likely related to cyber intrusion tools or operations.”OpenAI
Sources
- Date
- Feb 2025
- Actor
- Suspected VELVET CHOLLIMA (Kimsuky) and STARDUST CHOLLIMA (APT38) accounts (nation state), as attributed by OpenAI
- Nexus
- North Korea
- Targets
- Unspecified; interest in cryptocurrency and cyber intrusion tools
- What the AI did
- Malware dev, Recon
- Model as named
- ChatGPT
- Model basis
- The model's maker says so
- Autonomy
- Assisted a human
- Outcome
- OpenAI banned the accounts; staging URLs for unknown binaries it found were shared with the security community.