AI Darwin Awards to mock the year’s biggest failures in artificial intelligence

Published on 12/09/2025 - 8:30 GMT+2

A caller grant will observe bad, ill-conceived, aliases downright vulnerable uses of artificial intelligence (AI) — and its organisers are seeking nan internet’s input.

The AI Darwin Awards reward nan “visionaries” that “outsource our mediocre decision-making to machines”.

It has nary affiliation pinch nan Darwin Awards, a tongue-in-cheek grant that recognises group who “accidentally remov[e] their ain DNA” from nan cistron excavation by dying successful absurd ways.

To triumph 1 of nan AI-centred awards, nan nominated companies aliases group must person shown “spectacular misjudgement” pinch AI and “ignored evident informing signs” earlier their instrumentality aliases merchandise went out. 

Bonus points are fixed retired to AI deployments that made headlines, required emergency response, aliases “spawned a caller class of AI information research”.

“We’re not mocking AI itself — we’re celebrating nan humans who utilized it pinch each nan be aware of a toddler pinch a flamethrower,” an FAQ page astir nan awards reads.

Ironically, nan anonymous organisers said they will verify nominations partially done an AI fact-checking system, which intends they inquire aggregate ample connection models (LLMs) for illustration OpenAI’s ChatGPT, Anthropic’s Claude, and Google’s Gemini whether nan stories submitted are true.

The LLMs complaint a story’s truthfulness retired of 10, past nan administrators of nan tract mean nan scores pinch an AI calculator. If nan mean is supra five, nan communicative is considered “verified” and eligible for an AI Darwin Award. 

OpenAI, McDonald's among early nominees

One of nan approved nominations for nan first AI Darwin Awards is nan American accelerated nutrient concatenation McDonald's. 

The institution built an AI chatbot for occupation recruitment called “Olivia” that was safeguarded by an evident password: 123456, exposing a reported 64 cardinal people’s hiring information to hackers.

Another early nominee is OpenAI for nan motorboat of its latest chatbot exemplary GPT-5. French information intelligence Sergey Berezin claimed he sewage GPT-5 to unknowingly complete harmful requests “without ever seeing nonstop malicious instructions”.

The winners will beryllium wished by a nationalist ballot during nan period of January, pinch nan announcement expected successful February.

The only prize: “immortal nickname for their publication to humanity's knowing of really not to usage artificial intelligence,” nan organisers said.

The dream of nan awards is to service arsenic “cautionary tale[s]” for early decision-makers truthful they work together to trial AI systems earlier deploying them.