{
"$type": "site.standard.document",
"bskyPostRef": {
"cid": "bafyreifrckpw5dsiw37b7elzhwyqdvktngkfdvki3ifhdbzpojzbiyttoa",
"uri": "at://did:plc:ugfhomboczluft7hnwy7pkyw/app.bsky.feed.post/3mntbjrvjrkn2"
},
"coverImage": {
"$type": "blob",
"ref": {
"$link": "bafkreigaj4udjjg4nph4nlqkf2uyzxurbxfswvyr7xikxqahtjlncskjce"
},
"mimeType": "image/jpeg",
"size": 133772
},
"path": "/blog/indirect-prompt-injection/",
"publishedAt": "2026-06-08T00:00:00.000Z",
"site": "https://brave.com",
"tags": [
"AI news & features"
],
"textContent": "Instructions hidden in webpages and documents can hijack AI agents mid-task. Research shows indirect prompt injection hits cloud and on-device LLMs alike—local deployment doesn't eliminate the risk.",
"title": "Indirect Prompt Injection remains a fundamental security challenge for AI"
}