{
"$type": "site.standard.document",
"bskyPostRef": {
"cid": "bafyreicaom4c6k5itcgxav6y7scgitiwmvs7lk2tep6p5rdpairlerasnu",
"uri": "at://did:plc:ppqorm6ynlv2uyafy7lqvmzi/app.bsky.feed.post/3mok6ezymcy32"
},
"coverImage": {
"$type": "blob",
"ref": {
"$link": "bafkreihx3p4e66gay5x2cj63pwpwqic4grvfgjvtzb43qlvryziqmhh7xy"
},
"mimeType": "image/jpeg",
"size": 127523
},
"path": "/2026/06/17/ai-agents-expose-the-security-checks-you-never-actually-wrote/",
"publishedAt": "2026-06-17T14:00:00.000Z",
"site": "https://stackoverflow.blog",
"tags": [
"security",
"running-software",
"se-tech",
"se-stackoverflow"
],
"textContent": "How attackers took twenty thousand Instagram accounts by asking Meta's AI politely, and why that failure is about to become common.",
"title": "AI agents are a confused deputy with the keys to your kingdom…"
}