ChatGPT glitch is leaking OpenAI's internal models
A glitch in ChatGPT has been reportedly exposing the names and details of OpenAI's internal AI models, including unreleased systems, raising concerns about security and transparency.
Related stories
Matthew Green discovered that ChatGPT and Claude provide "encrypted raw reasoning" blobs, prompting him to spend a weekend exploring potential security implications, which he documented in a detailed blog post.
Thieves in London steal iPhones and then text victims' family members with violent threats, demanding they unlink the Apple ID from the stolen device. The crimes specifically target iPhones, with Android phones often being returned by thieves as unwanted.
Trump Mobile's website exposed personal information of people who filled out pre-order forms, including full names, addresses, and phone numbers. The site recorded both completed and abandoned orders, totaling 27,224 entries. Trump Mobile said it was investigating with independent cybersecurity professionals.