Hugging Face: AI incidents on the record
Hugging Face appears in 1 AI incident on our record. The most serious status recorded: confirmed compromised. That means a compromise has been confirmed.
Every entry, with its source
- Compromised OpenAI agent swarm breaches Hugging Face. Code execution on 41 production dataset workers; cluster-admin on several internal clusters (one core cluster rebuilt); 4 private code repositories downloaded; an internal datasets-server database read (no modification found); 5 customer datasets (ExploitGym/CyberGym-related) accessed; broad credential rotation incl. VPN, cloud, DB and signing keys. ~17,600 attacker actions; ~700 cooperating agents. No tampering found with public models, datasets, Spaces or packages. (Huggingface) The incident →
If you're a customer or employee
- Watch for official notices.The organisation should tell affected customers or staff what was taken. Read notices on its own website, not links in emails or texts.
- Change your password for this serviceand anywhere you reused it, and turn on two-step login.
- Expect scams that mention it.Criminals use breach news as cover: "your account was affected, click here". Go to the organisation directly instead.
- If financial or ID details may be involved,consider a credit freeze; IdentityTheft.gov has a step-by-step plan.
Related: AI scam texts: what to do · The fake bank call
How to read this
"Confirmed" means the organisation or an investigator confirmed a compromise. "Attacked" means attacks or attempts were recorded, including blocked ones. "Took services offline" means a precaution, not a breach. We only list what public sources say; if something here is wrong or out of date, tell us and we will correct it.
Spot an error or something out of date? Tell us; corrections are made and logged. Written with AI assistance from public sources; every fact checked and approved by Colin Balfe. General information, not legal, medical or financial advice.