diff --git a/lib/ai_moderation/spam_scanner.rb b/lib/ai_moderation/spam_scanner.rb
index e4ed348d..19e2433b 100644
--- a/lib/ai_moderation/spam_scanner.rb
+++ b/lib/ai_moderation/spam_scanner.rb
@@ -47,10 +47,20 @@ module DiscourseAi
         user = nil
         if SiteSetting.ai_spam_detection_user_id.present?
           user = User.find_by(id: SiteSetting.ai_spam_detection_user_id)
+          ensure_safe_flagging_user!(user)
         end
         user || Discourse.system_user
       end
 
+      def self.ensure_safe_flagging_user!(user)
+        # only do repair on bot users, if somehow it is set to a human skip repairs
+        return if !user.bot?
+        user.update!(silenced_till: nil) if user.silenced?
+        user.update!(trust_level: TrustLevel[4]) if user.trust_level != TrustLevel[4]
+        user.update!(suspended_till: nil, suspended_at: nil) if user.suspended?
+        user.update!(active: true) if !user.active?
+      end
+
       def self.after_cooked_post(post)
         return if !enabled?
         return if !should_scan_post?(post)
@@ -94,6 +104,9 @@ module DiscourseAi
         return false if !post.present?
         return false if post.user.trust_level > TrustLevel[1]
         return false if post.topic.private_message?
+        return false if post.user.bot?
+        return false if post.user.staff?
+
         if Post
              .where(user_id: post.user_id)
              .joins(:topic)
diff --git a/spec/lib/modules/ai_moderation/spam_scanner_spec.rb b/spec/lib/modules/ai_moderation/spam_scanner_spec.rb
index 97cd4dfa..52f303f4 100644
--- a/spec/lib/modules/ai_moderation/spam_scanner_spec.rb
+++ b/spec/lib/modules/ai_moderation/spam_scanner_spec.rb
@@ -49,6 +49,16 @@ RSpec.describe DiscourseAi::AiModeration::SpamScanner do
       expect(described_class.should_scan_post?(post)).to eq(false)
     end
 
+    it "returns false for bots" do
+      post.user.id = -100
+      expect(described_class.should_scan_post?(post)).to eq(false)
+    end
+
+    it "returns false for staff" do
+      post.user.moderator = true
+      expect(described_class.should_scan_post?(post)).to eq(false)
+    end
+
     it "returns false for users with many public posts" do
       Fabricate(:post, user: user, topic: topic)
       Fabricate(:post, user: user, topic: topic)
@@ -207,6 +217,26 @@ RSpec.describe DiscourseAi::AiModeration::SpamScanner do
     end
   end
 
+  it "unsilences flagging user if erronuously silenced" do
+    described_class.flagging_user.update!(silenced_till: 1.day.from_now)
+    expect(described_class.flagging_user.silenced?).to eq(false)
+  end
+
+  it "ensures flagging user is tl4" do
+    described_class.flagging_user.update!(trust_level: 0)
+    expect(described_class.flagging_user.trust_level).to eq(4)
+  end
+
+  it "unsuspends user if it was erronuously suspended" do
+    described_class.flagging_user.update!(suspended_till: 1.day.from_now, suspended_at: 1.day.ago)
+    expect(described_class.flagging_user.suspended?).to eq(false)
+  end
+
+  it "makes sure account is active" do
+    described_class.flagging_user.update!(active: false)
+    expect(described_class.flagging_user.active).to eq(true)
+  end
+
   describe "integration test" do
     fab!(:llm_model)
     let(:api_audit_log) { Fabricate(:api_audit_log) }
@@ -243,8 +273,13 @@ RSpec.describe DiscourseAi::AiModeration::SpamScanner do
     it "correctly handles spam scanning" do
       expect(described_class.flagging_user.id).not_to eq(Discourse.system_user.id)
 
-      # flag post for scanning
       post = post_with_uploaded_image
+      # this is surprising, core fabricator is not linking
+      # we need it linked so we scan uploads
+      post.link_post_uploads
+
+      expect(described_class.should_scan_post?(post)).to eq(true)
+      expect(post.upload_ids).to be_present
 
       described_class.new_post(post)