{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreieashmc2qzf4fv55amqzndcgzikj4gvfapdcdamum6v6kkjk5i7lu",
    "uri": "at://did:plc:oyu7kdlpydxm44ioz2z7kbhs/app.bsky.feed.post/3mj52xsjti5q2"
  },
  "coverImage": {
    "$type": "blob",
    "ref": {
      "$link": "bafkreiedopxnlbl2fntpa3jldoz37cyykaars4lol5xbaa5mftru477uae"
    },
    "mimeType": "image/png",
    "size": 35195
  },
  "path": "/2026/04/10/researchers-find-way-to-bypass-apples-on-device-llm-safeguards/",
  "publishedAt": "2026-04-10T09:49:02.000Z",
  "site": "https://dataconomy.com",
  "tags": [
    "Research",
    "Apple Intelligence"
  ],
  "textContent": "Researchers have found a way to bypass Apple's restrictions on its on-device LLM, successfully executing attacker-controlled actions through prompt injection.",
  "title": "Researchers find way to bypass Apple’s on-device LLM safeguards"
}