The activity taken by the ~1200 agents operated by OpenAI largely maps to techniques in the MITRE ATT&CK framework. This gist provides a brief mapping as observed by responders at Hugging Face and as reported by OpenAI.
| Report Quote | Tactic | Technique |
|---|---|---|
| Agents exploited a token-refresh vulnerability to obtain administrator-level Artifactory access | Privilege Escalation | [T1068 Exploitation for Privilege Escalati |