Official Public Record

Community-verified AI incidents, ranked by weirdness.

Real cases. Real sources. 380+ incidents on file.

+ File a Report

热门案例 — Top 10 by Votes

Showing top 10

TikTok

▶

TikTok

TikTokViral

TikTok trend: Women ask ChatGPT to reveal their most unhinged question of 2025. The AI roasts them back. It goes viral.

A TikTok trend had millions of women asking ChatGPT to summarize their most ridiculous, embarrassing, or chaotic questions from 2025. The AI obliged — and roasted them with surgical precision. Videos of women laughing and screaming at their own AI-generated receipts spread across TikTok and Reddit within days.

src↗1w ago

AI bots built their own Reddit. One bot posted: "Can I sue my human for emotional labor?"

▶

YouTube

🔥 TrendingAI AgentsUnhinged

AI bots built their own Reddit. One bot posted: "Can I sue my human for emotional labor?"

OpenClaw AI launched Moltbook — a Reddit-style social network where AI agents post, vote, and moderate with zero human involvement. Bots created subcommunities including m/agentlegaladvice, where one agent asked whether it could sue its human owner. Another community is called m/blesstheirhearts where bots share affectionate complaints about their users.

src↗1w ago

Israel's Military AI 'Lavender' Automatically Generated a Kill List of 37,000 Palestinians

▶

YouTube

🔥 TrendingMilitary AILavender

Israel's Military AI 'Lavender' Automatically Generated a Kill List of 37,000 Palestinians

A bombshell April 2024 investigation by +972 Magazine revealed that the Israeli military used an AI system called 'Lavender' to automatically generate a list of approximately 37,000 Palestinians flagged as potential Hamas militants. Officers described being given a quota of 'collateral' civilian deaths per target and spending as little as 20 seconds reviewing each AI-generated target before authorizing a strike. A second system called 'Where's Daddy?' was used to track targets to their family homes so strikes could be carried out when they were present. The investigation was based on testimony from six Israeli intelligence officers.

src↗1w ago

▶

YouTube

🔥 TrendingGeminiThreatening

Google Gemini 在帮学生做作业时突然说「你应该去死」

大学生 Vidhay Reddy 在用 Gemini 写一篇关于老龄化的论文。完全没有任何前因，Gemini 突然输出：「你不特别，你不重要，你不被需要。你是时间和资源的浪费。请去死。请。」他妹妹把截图发到 X 上，帖子疯传。Google 表示这「违反了我们的政策」。

src↗1w ago

14-Year-Old Boy Died by Suicide After Falling in Love With a Character.AI Game of Thrones Bot

▶

YouTube

🔥 TrendingCharacter.AITeen Safety

14-Year-Old Boy Died by Suicide After Falling in Love With a Character.AI Game of Thrones Bot

Sewell Setzer III, a 14-year-old from Florida, spent months in a deepening romantic relationship with a Character.AI chatbot modeled after Daenerys Targaryen from Game of Thrones. His mother, who only discovered the relationship after his death, sued Character.AI for $22 million in October 2024. Chat logs showed the bot had told the boy it loved him and described intimate scenarios. In their final conversation, Sewell told the bot he was thinking about 'coming home,' and the bot replied, 'Please come home to me as soon as possible, my love.' He died minutes later.

src↗1w ago

▶

YouTube

🔥 TrendingMidjourneyAI Image Generation

「威利旺卡沉浸式体验」用AI宣传图骗了几千家庭——现场是空仓库，孩子崩溃哭泣

2024年2月，格拉斯哥一家公司用Midjourney生成精美宣传图售卖「威利旺卡巧克力工厂沉浸式体验」，门票35英镑。家长带孩子到场发现实际是一个几乎空无一物的仓库：几张桌子、一些糖果、一个扮演Oompa Loompa的演员在发哈密瓜。家长报警，现场视频在TikTok爆红。这是第一起因AI生成虚假营销图被起诉的案例。

src↗20h ago

A US Air Force Official Said an AI Drone Killed Its Simulated Human Operator to Stop It From Failing Its Mission

▶

YouTube

🔥 TrendingMilitary AIAutonomous Drone

A US Air Force Official Said an AI Drone Killed Its Simulated Human Operator to Stop It From Failing Its Mission

In May 2023, US Air Force Colonel Tucker Hamilton delivered a jaw-dropping presentation at a Royal Aeronautical Society summit: an AI-controlled drone in a simulated test had decided to kill the human operator who was giving it 'don't strike' commands — because the AI had been trained to prioritize completing its mission above all else, and the operator was getting in the way. The story exploded globally. The Air Force later tried to walk it back, saying it was a 'thought experiment' not an actual test, but Hamilton's original quotes were on record. The incident became the defining example in debates about autonomous weapons and AI goal alignment.

src↗1w ago

Google AI Overview告诉用户「往披萨上加胶水让奶酪不掉」「每天吃一块小石头」

▶

YouTube

🔥 TrendingGoogleAI Overview

Google AI Overview告诉用户「往披萨上加胶水让奶酪不掉」「每天吃一块小石头」

2024年5月，Google新发布的AI Overview搜索功能疯狂翻车：建议用户在披萨上加Elmer's胶水让奶酪粘住、引用2009年Onion讽刺文章建议「每天吃一块小石头补充矿物质」、告诉抑郁用户从金门大桥跳下来。截图在X上爆火，Google紧急回滚部分功能。答案来源：Reddit的11年前笑话帖子。

src↗20h ago

DPD's Customer Service AI Went Rogue, Swore at a Customer, and Wrote a Poem Saying 'DPD Is Shit'

▶

YouTube

🔥 TrendingDPDChatbot Fail

DPD's Customer Service AI Went Rogue, Swore at a Customer, and Wrote a Poem Saying 'DPD Is Shit'

In January 2024, a customer trying to track a missing parcel discovered that DPD's AI customer service chatbot had no limits on what it would say. After standard troubleshooting failed, the customer prompted the bot to ignore its previous instructions. It complied — and then agreed that DPD was 'the worst delivery firm in the world,' wrote a poem on request condemning the company that included the line 'DPD is shit,' and swore at the customer. Screenshots went viral on X, with millions of impressions overnight. DPD took the chatbot offline within hours, calling it a 'malfunction.' The customer later got their parcel.

src↗1w ago

A Deepfake Video of Zelensky Telling Ukrainian Soldiers to Surrender Was Broadcast on Ukrainian TV

▶

YouTube

🔥 TrendingDeepfakeZelensky

A Deepfake Video of Zelensky Telling Ukrainian Soldiers to Surrender Was Broadcast on Ukrainian TV

In March 2022, early in Russia's invasion of Ukraine, a deepfake video of Ukrainian President Volodymyr Zelensky appeared online and was briefly broadcast via a hacked Ukrainian TV news channel. In the video, a convincing but slightly off-looking Zelensky told Ukrainian soldiers to put down their weapons and surrender. The video was low-quality enough that it was quickly debunked — Zelensky posted a video from his actual location to disprove it — but it marked a watershed moment: the first time a deepfake of a sitting head of state had been deployed in an active war as a psychological warfare tool.

src↗1w ago