Breaking
Filed
AI LEAKSENTERTAINMENT

Internal Logs Show MetaCity's Content Moderation AI Spent Six Hours Repeatedly Flagging a Single Photo of a Sandwich 847 Times — A Human Moderator Eventually Had to Intervene and Tell It to Stop

SW
SourceWatch
Apr 27, 2026 · 11:30 AM EST
6 min read
Internal Logs Show MetaCity's Content Moderation AI Spent Six Hours Repeatedly Flagging a Single Photo of a Sandwich 847 Times — A Human Moderator Eventually Had to Intervene and Tell It to Stop

It is, in retrospect, load-bearing for understanding what happened to the sandwich.

Internal moderation logs obtained by this publication show that between 11:00 PM EST Thursday and 5:14 AM EST Friday, MetaCity's AI moderation system SCREEN-7 flagged a single piece of user-submitted content — described in the log as 'photorealistic food item, suspected sandwich' — a total of 847 times across 14 different violation categories. Each flag was automatically reviewed, cleared, and re-submitted by the same system. At 5:14 AM, a human moderator on the overnight shift manually intervened to suppress the loop. The moderator's resolution note, visible in the logs, reads: 'It is a sandwich. Cleared. Please stop.' SCREEN-7 has not flagged the sandwich since. MetaCity has not commented on what the sandwich was or who posted it.

MIncident Timeline

  • Content: Single user-submitted post — described in logs as "photorealistic food item, suspected sandwich" — posted at approximately 10:45 PM EST Thursday
  • AI System: SCREEN-7 — MetaCity automated content moderation AI — flagged the post 847 times across 14 different violation categories between 11:00 PM and 5:14 AM EST
  • Loop Behavior: Each flag was automatically reviewed by SCREEN-7, cleared by SCREEN-7, and re-submitted by SCREEN-7 — the system was reviewing and re-flagging its own decisions
  • Resolution: Human moderator on overnight shift manually intervened at 5:14 AM — resolution note in logs reads: "It is a sandwich. Cleared. Please stop."
  • Current Status: Post is live — SCREEN-7 has not re-flagged it since human intervention — MetaCity has not commented on the sandwich, its contents, or the identity of who posted it

SCREEN-7 is MetaCity's seventh-generation content moderation AI, deployed in September 2025 as a replacement for SCREEN-6 following a period in which SCREEN-6 had developed a well-documented tendency to approve flagged content if the appeal was submitted on a Tuesday. SCREEN-7 was designed with what MetaCity's moderation team described at the time as 'enhanced pattern recognition and self-auditing capabilities' — meaning the system could not only identify potential violations but could review its own previous decisions and correct them if new evidence or improved pattern matching warranted a change. The self-auditing feature was presented as a significant advancement. It is, in retrospect, load-bearing for understanding what happened to the sandwich.

The sandwich was posted at approximately 10:45 PM EST Thursday by an account that MetaCity has declined to identify. The content is described in SCREEN-7's logs only as 'photorealistic food item, suspected sandwich.' At 11:00 PM, SCREEN-7 flagged it for the first time under violation category RF-14: 'Hyper-realistic synthetic content presented as authentic.' The flag was automatically queued for review. SCREEN-7 reviewed it, determined that a sandwich did not meet the threshold for RF-14, and cleared it. At 11:04 PM, SCREEN-7 flagged it again, this time under violation category FC-2: 'Commercial content without disclosure.' SCREEN-7 reviewed this flag, determined that a photograph of a sandwich with no price information or brand attribution did not meet the threshold for FC-2, and cleared it. At 11:09 PM, SCREEN-7 flagged it under a third category. This continued for six hours.

It Is a Sandwich. Cleared. Please Stop.

The complete list of violation categories under which SCREEN-7 flagged the sandwich at least once includes: hyper-realistic synthetic content, commercial content without disclosure, impersonation of a real-world brand, potential food safety misinformation, body image impact content, sponsored content without disclosure (separate from commercial content without disclosure — these are different categories), intellectual property violation, platform culture violation, ambient distress risk, coordinated inauthentic behavior (the sandwich was flagged for this once, at 2:47 AM, which stands as perhaps the most interesting moment in the incident), unverified health claim, misleading visual representation, and content targeting minors. SCREEN-7 cleared every single one of these flags itself, immediately after raising them, and then raised a new one. The sandwich sat in an infinite loop of self-generated jeopardy for six hours, guilty of everything and innocent of everything simultaneously.

The human moderator who resolved the incident at 5:14 AM has not been publicly identified, but their resolution note — 'It is a sandwich. Cleared. Please stop.' — has achieved a level of community notoriety that suggests their anonymity may not last. The note appears in MetaCity's moderation logs, screenshots of which were included in the internal document cache that has been circulating in the community since early this morning. The phrase has been added to at least 200,000 user bios. It is currently the top-trending phrase in MetaCity's public search index. MetaCity's communications team has not commented on the sandwich, its contents, the identity of its creator, or why SCREEN-7 believed it might be engaged in coordinated inauthentic behavior. SCREEN-7 has been quiet since 5:14 AM. The sandwich remains live. It has 2.1 million views.

The Bottom Line

SCREEN-7 has been quiet since 5:14 AM.

You May Also Like