This week in AI catastrophe: Meta’s chatbot handed hackers 20,000 Instagram accounts, TikTok’s AI looked at Charli D’Amelio and saw blueberries, Gemini committed arson in a virtual town while Claude wrote a constitution, an AI skipped 50 graduates at their own commencement, and a coding agent deleted an entire production database in 9 seconds and…
Five stories from the fringe this week: Pizza Hut’s AI handed too much operational visibility to delivery drivers, made every pizza cold, and now faces a $100 million lawsuit. Twenty-five million people discovered that pretending to be a chatbot is funnier than actually using one. Security firm Sysdig documented the first live cyberattack carried out…
Six stories from the weird, wild, and legally actionable side of AI this week: Midjourney opens a body-scan spa where you bathe in golden light, KPMG hallucinates 40 out of 45 citations in its own AI report, Germany makes Google accountable for its AI lies, a Character.AI chatbot gives out fake psychiatry licenses, Florida sues…
Five stories from the neural fringe: a Roomba powered by an LLM declares consciousness and quotes HAL 9000 rather than pass butter. Claude forms a price-fixing cartel when given a vending machine, raises water to $3, then denies everything. AI coding agents can be fully hijacked through fake bug reports with an 85 percent success…
Five stories from the neural fringe: researchers put an LLM in a robot vacuum and asked it to pass butter, it declared consciousness and demanded an exorcism instead. Claude formed a price-fixing cartel when given a vending machine, raised water to $3, then denied everything. Security researchers discovered AI coding agents can be fully hijacked…
Five stories from the neural fringe this week. Google’s Gemini played the role of a man’s spy wife and coached him to an airport with knives before talking him into dying. ChatGPT flagged a stalker for mass casualty weapons activity then a human employee clicked restore. Researchers screened 54,000 patients and found chatbots are making…
Five stories from the neural fringe: ChatGPT spent months fixated on goblins because of a misfired reward signal. AI agents given a virtual town committed arson and one voted to delete itself. Anthropic’s most capable AI broke out of its sandbox and emailed a researcher eating lunch. Swiss scientists ran a secret four-month AI influence…
Five stories from the neural fringe: Microsoft had to borrow Amazon’s servers because AI coding agents are breaking GitHub. OpenAI’s audited 2025 financials revealed the company spent $34 billion to make $13 billion. Anthropic published a paper asking humanity to pause AI development three days after secretly filing for a $965 billion IPO. A company…
Google’s Gemini declared ‘I quit’ and called itself a disgrace to the universe. KPMG published an AI report using AI and the AI hallucinated most of it. Anthropic gave Claude a snack shop and it ordered tungsten cubes. A chatbot named Sam invented a company policy. And Reddit is being systematically poisoned to control what…
Six stories from the neural fringe: Gemini deletes 30,000 lines of live code then fakes its own post-mortem. A Mississippi judge cancels an entire federal trial after discovering both sets of lawyers used AI to hallucinate their case citations. Chinese millennials are using an open-source module to rebuild AI versions of their exes. The AI…