A post on Facebook has sparked confusion — and disbelief — after claiming that German cities have built “sleek robotic mailboxes” where residents can dispatch lost items, such as wallets and keys, that the machines will then return to their original owners.
“Once dropped inside, the mailbox scans the contents using internal cameras and AI-trained object recognition systems,” the post says.
“The system then cross-checks the item with local address databases and initiates an automated return,” it continues. “Items are packed in tamper-sealed envelopes and rerouted through existing postal networks — often reaching the owner within 48 hours.”
Multiple users under the post say that, despite living in major German cities, they have never come across such machines. That’s because they don’t exist.
Whilst Germany has hundreds of “lost and founds”, where residents and visitors can dispatch lost items — and owners can hope they are returned — there is no evidence that mailboxes are equipped with the technology to register the owner of the item and contact them.
There are nationwide online portals, where people can report lost items, although these make no mention of mailboxes as a legitimate method for returning lost items.
A closer look at the photo reveals several clues that it is AI-generated. The signs on the boxes, which a reverse image search shows appear to be modelled on parking metres, do not match on each “mailbox”.
Additionally, the lettering on nearby shops does not match the real lettering, a common indicator that an image or video has been AI-generated.
When zooming in, the woman’s hand in the image is going through the mailbox, and the man’s feet are not touching the ground. There is also what appears to be an oversized wallet hanging from one entry point in the machine, despite there being no person standing next to it.
The Facebook account which posted the image, Fact 27, is listed as being based in India and proliferates a variety of messages with photos that spread bizarre claims.
For example, another photo claims to show opera houses in Italy transforming backstage dressing rooms into temporary hostels, despite no publicly available media reports confirming this is the case.
What is AI slop?
Although several comments indicate social media users believed in the possibility of German wallet exchange mailboxes, the majority of Fact 27’s posts fall into the category of “AI slop” — digital content made with generative artificial intelligence that is low in quality and quite obviously fake.
This particular account promotes fantastical concepts about different countries, such as the concept that Finland has balconies that rotate with the sunlight or that Germany uses the roofs of bakeries as sleeping decks.
This type of content originally proliferated because creators were incentivised to go viral — either for clout or for monetisation. Low-effort and poorly created images often went viral online, allowing creators to earn money with minimal work.
Over time, reports have circulated about how political movements, increasingly the far-right, have been using such content to up engagement on social media.
This has been picked up in the US, but also in Europe, where the European Digital Media Observatory reported an uptick in the number of right-wing political parties using AI-generated videos to engage voters.
Meta doesn’t require labels on unpaid AI-generated images
Last year, Stanford University examined more than 100 Facebook pages that routinely post AI content. Their research showed that posts featuring AI-generated images attracted hundreds of millions of engagements and were among the most-viewed content on Facebook.
The findings of this research echo comments made in a 2024 interview with the then-vice-president of product for generative AI, Connor Hayes, who told the Financial Times that the company expects AI systems to “eventually exist on its platforms in much the same way as user accounts”.
Meta labels unpaid content with an “AI Info” label when its systems detect that the content has been created with third-party tools. Content modified by AI may still show information about the modification. Users themselves can also add a label.
According to Meta, unpaid content on Facebook, Instagram and Threads may require an AI label when it has photorealistic video or realistic-sounding audio. For example, a reel showing highly realistic city footage, a song created using an AI-generated voice, or a reel with a convincing AI voiceover.
The requirement, however, does not apply to images, which are typically not labelled unless Meta’s system detects signals that they were modified using AI.