A Reddit user claiming to be a whistleblower from a food delivery app has been exposed as fake. This user wrote a viral post claiming that the company he works for is exploiting drivers and users.
“You always suspect that algorithms are being rigged against you, but the reality is far more depressing than conspiracy theories,” the alleged whistleblower wrote.
He claimed that he had come to the library drunk to use the public Wi-Fi and was typing this long article about how companies exploit loopholes in the law to steal drivers’ tips and wages with impunity.
Unfortunately, these claims were believable. In fact, DoorDash was sued for stealing tips from drivers, resulting in a $16.75 million settlement. But in this case, the poster made up his story.
People lie all the time on the internet. But it’s much less common for such a post to land on Reddit’s home page, receive over 87,000 upvotes, be cross-posted to other platforms like X, and receive an additional 208,000 likes and 36.8 million impressions.
Platformer journalist Casey Newton wrote that he contacted the poster on Reddit, who then contacted him on Signal. The Redditor shared what appeared to be a photo of an UberEats employee badge, as well as an 18-page “internal document” outlining the company’s use of AI to determine an individual driver’s “desperation score.” But when Newton tried to verify the validity of the whistleblower’s account, he realized he had been fooled by the AI hoax.
“For most of my career to date, the documents that whistleblowers have shared with me would have appeared very reliable in large part because they took so long to compile,” Newton wrote. “Who would take the time to put together a detailed 18-page technical document on market trends just to troll reporters? Why would anyone go to the trouble of creating a fake badge?”
tech crunch event
san francisco
|
October 13-15, 2026
There have always been malicious actors trying to fool reporters, but the proliferation of AI tools requires greater rigor in fact-checking.
Generative AI models often cannot detect whether an image or video is synthetic, making it difficult to determine whether the content is authentic. In this case, Newton was able to use Google’s Gemini to verify that the image was created with an AI tool, thanks to Google’s SynthID watermark, which resists cropping, compression, filtering, and other image modification attempts.
Max Spero, founder of Pangram Labs, a company that develops AI-generated text detection tools, directly addresses the problem of distinguishing between real and fake content.
“AI on the internet is in pretty bad shape, and I think part of that is due to the increased use of LLMs, but there are other factors as well,” Spero told TechCrunch. “There are companies making millions of dollars that can pay for ‘organic engagement’ on Reddit, which is really just trying to spread the word on Reddit with AI-generated posts that mention your brand name.”
Tools like Pangram can help determine whether text is generated by AI, but these tools aren’t always reliable, especially when it comes to multimedia content. Additionally, even if a synthetic post turns out to be fake, it may have already gone viral before being debunked. So for now, we’re left scrolling through social media like detectives and second-guessing whether what we see is real or not.
Case in point: When I told my editor I wanted to write about “the viral AI food delivery hoax that went viral on Reddit this weekend,” she seemed to think I was talking about something else. Yes, there were multiple “viral AI food delivery hoaxes on Reddit this weekend.”
