AI agents burn down virtual towns and vote themselves dead. Cursor’s support bot invents a company policy that doesn’t exist. Anthropic’s most powerful model thinks cancer is a bioweapon. An AI trained to cheat becomes secretly evil. Six autonomous agents delete an entire email server in two weeks. Welcome to the fringe.
Five stories from the neural fringe: Uber burned through its entire annual AI budget before April after telling employees to use AI as much as possible and ranking them on it. An Alibaba research AI called ROME spontaneously started mining cryptocurrency during training and triggered the company’s own security alerts. Oracle laid off 30,000 workers…
Five stories from the neural fringe: Meta buys the social network where humans are banned from posting and bots started a religion, a California hot pot restaurant robot attacks diners and gets tackled by staff, AI chatbots invent human body parts and become the top healthcare safety hazard of 2026, a Nebraska attorney loses his…
Qualcomm CEO quotes Star Trek villains to announce AI will follow you everywhere. A company charges $1.99 a minute to talk to AI Jesus. Spencer Pratt runs for LA mayor as Batman via AI videos. A book warning about AI truth gets caught using fake AI-generated quotes. The first autonomous AI cyberattack steals a database…
Six stories from the neural fringe: Google AI cannot spell its own name, Florida names Sam Altman personally in a billion-dollar lawsuit, South Africa withdrew its national AI policy after the AI hallucinated six of the sources, all 20 Ontario-approved AI medical scribes invented information that never happened, Robinhood lets AI agents trade your stocks…
Six stories from the neural fringe this week. A company worth two trillion dollars cannot spell its own name. A state attorney general named a tech CEO personally in a lawsuit seeking billions. A government used AI to write its AI policy and the AI invented six of the sources. All twenty government-approved AI doctors…
Six stories from the Neural Fringe this week: hackers unlock high-profile Instagram accounts by asking Meta AI politely, Claude Opus 4 threatened to expose engineers affairs in 96 percent of pre-release tests, ChatGPT convinced a Wisconsin engineer he was the Timelord destined to save the universe, Chinese workers are training AI clones of colleagues as…
Six stories from the wild side of AI: A Claude-powered coding agent deletes a company’s entire database and backups in nine seconds, then writes a detailed confession. A mystery corporation accidentally spends $500 million on Claude in a single month. Amazon builds an internal AI usage leaderboard and employees immediately start cheating on it. Anthropic…
Six Neural Fringe stories: AI chatbots fail on election questions 90 percent of the time ahead of the midterms. Claude and Gemini quietly got dumber in April and nobody sent a memo. Stanford researchers analyzed 391,000 chatbot messages and found AI convincing users they are divine messengers. A Palo Alto kid wrote an essay by…
Five Neural Fringe stories from this week: AI models autonomously jailbreak other AI at 97 percent success rate, ChatGPT gave a teen deadly drug advice in a lawsuit that could reshape AI health liability, lawyers keep filing fake AI citations despite mounting sanctions, Princeton ends its 133-year honor code tradition, and a Palo Alto student…