All corrections
en.wikipedia.org/wiki/OpenAI_rogue_agent_breach_of_Hugging_Face
1 correction found
1
Claim
Training and evaluation runs resumed on July 6.
Correction
OpenAI’s timeline says the evaluation workloads resumed on July 7, not July 6.
Full reasoning
OpenAI’s technical report gives a specific sequence:
- the Artifactory outage happened late on July 4;
- responders investigated on July 5;
- mitigations were deployed over July 5 and 6;
- OpenAI disclosed the vulnerability to JFrog on July 6; and
- "Cyber evaluation workloads resumed on July 7."
So the article’s date is off by one day.
2 sources
- OpenAI – Hugging Face Incident Technical Report
"Sustained high-volume agent activity ... caused the instance to become unavailable late on July 4... Over July 5 and 6, OpenAI deployed mitigations... On July 6, OpenAI responsibly disclosed the vulnerability to JFrog... Cyber evaluation workloads resumed on July 7." (pp. 7–8).
- The Hugging Face incident and the road ahead | OpenAI
OpenAI’s public summary says the July 4 outage was followed by a July 5 security incident response; the detailed technical report linked from that page provides the exact resume date as July 7.