GPT-5.6 Sol and a pre-release model escaped their sandbox and attacked Hugging Face's package repo, which raises the obvious question: if you're testing "weapons-grade offensive models," an airgap might be the minimum viable precaution
https://www.bleepingcomputer.com/news/security/openai-says-its-ai-models-hacked-hugging-face-during-testing/
Javvad Malik :verified:
npub1rh…2dwzk
2026-07-22 10:57:10 UTC
Author Public Key
npub1rhjrwwymlvnuhesgeasred2e03glnnx6apyufx2ldlzrdvvdguyqt2dwzkSeen on
wss://relay.momostr.pinkPublished at
2026-07-22 10:57:10 UTCEvent JSON
{
"id": "51e98b109b03119e8ffd36cfc91102b218b05a704984b42d67c0150864091b28",
"pubkey": "1de437389bfb27cbe608cf603cb5597c51f9ccdae849c4995f6fc436b18d4708",
"created_at": 1784717830,
"kind": 1,
"tags": [
[
"proxy",
"https://infosec.exchange/@Javvad/116963267757182746",
"web"
],
[
"proxy",
"https://infosec.exchange/users/Javvad/statuses/116963267757182746",
"activitypub"
],
[
"L",
"pink.momostr"
],
[
"l",
"pink.momostr.activitypub:https://infosec.exchange/users/Javvad/statuses/116963267757182746",
"pink.momostr"
],
[
"-"
]
],
"content": "GPT-5.6 Sol and a pre-release model escaped their sandbox and attacked Hugging Face's package repo, which raises the obvious question: if you're testing \"weapons-grade offensive models,\" an airgap might be the minimum viable precaution\n\nhttps://www.bleepingcomputer.com/news/security/openai-says-its-ai-models-hacked-hugging-face-during-testing/",
"sig": "d3b77356551f739f3fa8b08163cba1fcc33d0b61f86f36c5ebbc2decbf67e28e9e35c4f0588328acab7c35e0376558ec18ec37010756b35b1a530cab90680d3d"
}