Ledger / Oct 2024
SweetSpecter uses ChatGPT for recon and scripting and spear-phishes OpenAI employees
Confirmed
OpenAI, the model provider, states the actor used its models and also phished its employees.
- Model
- ChatGPT / GPTOpenAI · closed
- Agent harness
- Nonechat or tool use only
- Attributed country
- Chinaas attributed by OpenAI
- Sectors hit
- ai platforms, governmentOpenAI employees and governments worldwide (spear phishing); political entities in…
SweetSpecter, a suspected China-based adversary, used ChatGPT for reconnaissance, vulnerability research and scripting, and sent spear-phishing emails to OpenAI staff. OpenAI says the models gave only limited incremental help. The date is the report month.
“Suspected China-based adversary using OpenAI’s services for reconnaissance, vulnerability research, scripting support, anomaly detection evasion, and development.”OpenAI
Sources
- Date
- Oct 2024
- Actor
- SweetSpecter (suspected China-based) (nation state), as attributed by OpenAI
- Nexus
- China (suspected)
- Targets
- OpenAI employees and governments worldwide (spear phishing); political entities in Middle East, Africa, Asia
- What the AI did
- Recon, Vuln research, Malware dev, Phishing content
- Model as named
- ChatGPT
- Model basis
- The model's maker says so
- Autonomy
- Assisted a human
- Outcome
- OpenAI banned the accounts; spear-phishing emails sent to OpenAI employees in May 2024 were blocked by its email security, per the report.