diff --git a/nobilis b/nobilis
index 046fbb4..ca29a63 160000
--- a/nobilis
+++ b/nobilis
@@ -1 +1 @@
-Subproject commit 046fbb4850b95a3f5ed0759038792eb5e2a7be0c
+Subproject commit ca29a638830673ce0277249f4e934a927c1152d0
diff --git a/src/renderer/src/lib/format.test.ts b/src/renderer/src/lib/format.test.ts
index 42585f2..a954410 100644
--- a/src/renderer/src/lib/format.test.ts
+++ b/src/renderer/src/lib/format.test.ts
@@ -9,6 +9,7 @@ import {
   extractMedia,
   extractQuoteBlocks,
   formatMessage,
+  isForumUpload,
   isGifPage,
   hostnameOf,
   ircFormat,
@@ -286,3 +287,23 @@ describe('links to the pages GIFs are shared from', () => {
     expect(media.map((m) => m.kind)).toEqual(['video', 'image'])
   })
 })
+
+describe('the forum\'s own files', () => {
+  const picture = 'https://uploads.kiwifarms.st/data/attachments/9671/9671535-a2e5ae0d55aadd338162770b3fa79f97-alt.avif?hash=-d_TarFo41'
+
+  it('are known by where they are', () => {
+    expect(isForumUpload(picture)).toBe(true)
+    expect(isForumUpload('https://kiwifarms.st/data/video/9667/9667943-abc.mp4?hash=x')).toBe(true)
+    expect(isForumUpload('https://kiwifarms.st/threads/a-thread.123/')).toBe(false)
+    expect(isForumUpload('https://example.com/data/attachments/1/2-a.avif')).toBe(false)
+  })
+
+  it('are not loaded by the window, which would ask for them from the reader\'s own address', () => {
+    expect(extractMedia(`[img]${picture}[/img]`, {})).toEqual([])
+    expect(extractMedia(`look ${picture}`, { contentSniffing: true, onNeedSniff: () => { throw new Error('asked') } })).toEqual([])
+  })
+
+  it('leave other pictures and an AVIF anywhere else as they were', () => {
+    expect(extractMedia('https://i.ibb.co/JRkN8BFv/photo.avif', {}).map((m) => m.kind)).toEqual(['image'])
+  })
+})
diff --git a/src/renderer/src/lib/format.ts b/src/renderer/src/lib/format.ts
index 998bb51..0c49354 100644
--- a/src/renderer/src/lib/format.ts
+++ b/src/renderer/src/lib/format.ts
@@ -594,6 +594,27 @@ const VIDEO_EXT_RE = /\.(mp4|webm|mov|mkv|ogv)(\?\S*)?$/i
  */
 const GIF_PAGE_RE = /^https?:\/\/(?:www\.)?(?:tenor\.com\/(?:[a-z]{2}(?:-[A-Za-z]{2})?\/)?view\/|giphy\.com\/(?:gifs|embed)\/|klipy\.com\/(?:gifs?|clips?|stickers?|memes?)\/)/i
 
+/**
+ * One of the forum's own files: what its uploads host serves, a picture or a
+ * video.
+ *
+ * Not loaded by the window. The daemon fetches these through the account's own
+ * route and puts the file on the message, which is what draws; left to the
+ * window they would be asked for from the reader's own address - which the host
+ * turns away, and which for somebody on Tor is the leak routing is for. Between
+ * the message arriving and the file being kept the link is simply a link, and if
+ * it cannot be kept it stays one.
+ *
+ * The onion address is here as the clearnet one because the text has been
+ * through `clearnetLinks` by the time it reaches this.
+ */
+const FORUM_UPLOAD_RE = /^https?:\/\/(?:uploads\.|www\.)?kiwifarms\.st\/data\/(?:attachments|video)\//i
+
+/** Whether this is a link to one of the forum's own files. */
+export function isForumUpload(url: string): boolean {
+  return FORUM_UPLOAD_RE.test(url)
+}
+
 /** Whether this is a link to a GIF's page on one of those sites. */
 export function isGifPage(url: string): boolean {
   return GIF_PAGE_RE.test(url)
@@ -668,7 +689,7 @@ export function extractMedia(
     const url = raw.replace(/[),.;!?]+$/, '')
     const yt = youtubeId(url)
     const identity = yt ? `yt:${yt}` : url
-    if (seen.has(identity) || targets.has(url)) continue
+    if (seen.has(identity) || targets.has(url) || isForumUpload(url)) continue
 
     if (yt) {
       seen.add(identity)