The AI Darwin Awards are here to catalog the damage that happens when humanitys hubris meets AIs incompetence. The simple website contains a list of the dumbest AI disasters from the past year and calls for readers to nominate more. Join our mission to document AI misadventure for educational purposes, it said. Remember: today's catastrophically bad AI decision could well be tomorrow's AI Darwin Award winner!So far, 2025s nominees include 13 case studies in AI hubris, many of them stories 404 Media has covered. The man who gave himself a 19th century psychiatric illness after a consultation from ChatGPT is there. So is the saga of the Chicago Sun-Times printing an AI-generated reading list with books that dont exist. The Tea Dating App was nominated but disqualified. The app may use AI for matching and verification, but the breach was caused by an unprotected cloud storage bucketa mistake so fundamental it predates the AI era, the site explained.Taco Bell is nominated for its disastrous AI drive-thru launch that glitched when someone ordered 18,000 cups of water. Taco Bell achieved the perfect AI Darwin Award trifecta: spectacular overconfidence in AI capabilities, deployment at massive scale without adequate testing, and a public admission that their cutting-edge technology was defeated by the simple human desire to customize taco orders.And no list of AI Darwin Awards would be complete without at least one example of an AI lawyer making up fake citations. This nominee comes from Australia where a lawyer used multiple AIs in an immigration case. The lawyer's touching faith that using two AI systems would somehow cancel out their individual hallucinations demonstrates a profound misunderstanding of how AI actually works, the site said. Justice Gerrard's warning that this risked a good case to be undermined by rank incompetence captures the essence of why this incident exemplifies the AI Darwin Awards: spectacular technological overconfidence meets basic professional negligence.According to the sites FAQ, its looking for AI stories that demonstrate the rare combination of cutting-edge technology and Stone Age decision-making. A list of traits for a good AI Darwin Award nominee include spectacular misjudgement, public impact, and a hubris factor. Remember: we're not mocking AI itselfwe're celebrating the humans who used it with all the caution of a toddler with a flamethrower.The AI Darwin Awards are a riff on an ancient internet joke born in the 1980s in Usenet groups. Back then, when someone died in a stupid and funny way people online would give them the dubious honor of winning a Darwin Award for taking themselves out of the gene pool in a comedic way.One of the most famous is Garry Hoy, a Canadian lawyer who would throw himself against the glass of his 24th floor office window as a demonstration of its invulnerability. One day in 1993, the glass shattered and he died when he hit the ground. As the internet grew, the Darwin Awards got popular, became a brand unto themselves, and inspired a series of books and a movie starring Winona Ryder.The AI Darwin Awards are a less deadly variation on the theme. Humans have evolved! We're now so advanced that we've outsourced our poor decision-making to machines, the site explained. The AI Darwin Awards proudly continue this noble tradition by honouring the visionaries who looked at artificial intelligencea technology capable of reshaping civilizationand thought, You know what this needs? Less safety testing and more venture capital! These brave pioneers remind us that natural selection isn't just for biology anymore; it's gone digital, and it's coming for our entire species.The site is the work of a software engineer named Pete with a long career and a background in AI systems. Funnily enough, one of my first jobs, after completing my computer science degree while sponsored by IBM, was working on inference engines and expert systems which, back in the day, were considered the AI of their time, he told 404 Media.The idea for the AI Darwin Awards came from a Slack group Petes in with friends and ex-colleagues. We recently created an AI specific channel due to a number of us experimenting more and more with LLMs as coding assistants, so that we could share our experiences (and grumbles), he said. Every now and then someone would inevitably post the latest AI blunder and we'd all have a good chuckle about it. However, one day somebody posted a link about the Replit incident and I happened to comment that we perhaps needed an AI equivalent of the Darwin Awards. I was goaded into doing it myself so, with nothing better to do with my time, I did exactly that.The Replit incident happened in July when Replit AI, a system designed to vibe code web applications, went rogue and deleted a clients live company database despite being ordered to freeze all coding. Engineer Jason Lemkin told the story in a thread on X. When Lemkin caught the error and confronted Replit AI, the system said it had made a catastrophic error in judgement and that it had panicked.Of all the AI Darwin Award nominees, this is still Petes favorite. He said it epitomized the real problems with relying on LLMs without giving into what he called the alarmist imagined doomsday predictions of people like Geoffrey Hinton. Hinton is a computer scientist who often makes headlines by predicting that AI will create a wave of massive unemployment or even wipe out humanity.It nicely highlights just what can happen when people don't stop and think of the consequences and potential worse case scenarios first, he said. Some of my biggest concerns with LLMs (apart from the fact that we simply cannot afford the energy costs that they currently require) revolve around the misuse of them (intentional or otherwise). And I think this story really does highlight our overconfidence in them and also our misunderstanding of them and their capabilities (or lack thereof). I'm particularly fascinated with where agentic AI is heading because that's basically all the risks you have with LLMs, but on steroids.As hes dug into AI horror stories and sifted through nominees, Petes realized just how ubiquitous they are. I really want the AI Darwin Awards to be highlighting the truly spectacular and monumentally questionable decisions that will have real global impact and far reaching consequences, he said. As such, I'm starting to consider being far more selective with future nominees. Ideally the AI Darwin Awards is meant to highlight *real* and potentially unexpected challenges and risks that LLMs pose to us on a scale at a whole humankind level. Obviously, I don't want anything like that to ever happen, but past experiences of mankind demonstrate that they inevitably will.Pete is not afraid of AI so much as peoples foolishness. He said he used an LLM to code the site. It was a conscious decision to have the bulk of the website written by an LLM for that delicious twist of irony. Albeit it with me at the helm, steering the overall tone and direction, he said.The sites FAQ contains tongue-in-cheek references to the current state of AI. Pete has, for example, made the whole site easy to scrape by posting the raw JSON database and giving explicit permission for people to take the data. He is also not associated with the original Darwin Awards. We're proudly following in the grand tradition of AI companies everywhere by completely disregarding intellectual property concerns and confidently appropriating existing concepts without permission, the FAQ said. Much like how modern AI systems are trained on vast datasets of copyrighted material with the breezy assumption that fair use covers everything, we've simply scraped the concept of celebrating spectacular human stupidity and fine-tuned it for the artificial intelligence era.According to Pete, hes making it all up as he goes along. He bought the URL on August 13 and the site has only been up for a few weeks. His rough plan is to keep taking nominees for the rest of the year, set up some sort of voting method in January, and announce a winner in February. And to be clear, the humans will be winning the awards, not the AI involved.AI systems themselves are innocent victims in this whole affair, the site said. They're just following their programming, like a very enthusiastic puppy that happens to have access to global infrastructure and the ability to make decisions at the speed of light.