Topic
Covers risks of AI systems collecting, storing, and reusing personal data, including leakage or exposure to human reviewers across chatbots and agents.
Researchers say AI companies have at least twice in five weeks reached answers to questions they had long been working on, Nature reported. Some now limit their AI use, though one AI-safety scientist expects models to keep getting there first with or without uploaded manuscripts.
Reality
DeepSeek released its open-source Harness agent runtime as a Windows and macOS desktop app on September 30, so users no longer need a command line to run it. The agent edits local files and runs commands, so what each user exposes to it decides what it can touch and which provider sees it.
Perspective Coverage
3 publishersReality
Anthropic is asking Claude users to share voice chats for training, extending a 2025 text policy that keeps opted-in data up to five years. Teams with staff on consumer seats now have a second data setting to check before anyone talks to Claude.
Reality
Contractors hired to grade Copilot's image edits see each user's original photo, faces uncensored, beside the request and two AI outputs, 404 Media reported. Microsoft did not say whether users can keep their uploads out of those reviews.
Reality
Meta limits its Muse AI agent to users 18 and over while selling it through Jolly, a dress-up mascot, and a planned Tamagotchi-style device. Fairplay says the character appeals to preschoolers, and Meta answers with extra age checks it has not described.
Reality
The Verge's hands-on found Muse cleaning an inbox and buying tank tops as advertised, but the consent screen only covers the accounts a user points it at, and the assistant volunteered data from ones nobody asked about.
Perspective Coverage
11 publishersReality
Sensor Tower counted more than 900,000 downloads of Meta's Muse in its first week. Wired's reviewer found an agent that browses competently while its ideas tab keeps pitching a live link to the bank.
Reality
404 Media obtained instruction guides, Slack channels and real prompts from OpenAI's human review program. Reviewers see full chats and a summary of what the user asked for before, with the username stripped.
Reality
The tool is free and needs no log-in, which makes a pre-prompt scrub adoptable in seconds and impossible to evidence, so the real decision is where the step sits in the workflow rather than whether you announce it.
Reality
Portable Computer keeps agent work on hardware you own and asks permission before any single step escalates. The entry fee is 24GB of VRAM and a paid subscription.
Reality
Charleston County wrote its own AI policy because South Carolina has not published guidance, and phase two is training. The demos that teach students to distrust models expire as the models improve.
Reality