Will AI Kill Us? | Daily AI News.

OpenAI Calls In Referees for Its Own Homework

OpenAI says it has three new AI-assisted results in math, computer science, and physics — but this time it's bringing in outside academic 'shepherds' to check the work be

A Wall Street Upgrade for a Robot Company — and Why It Doesn't Tell Us Much

JPMorgan just made a big bullish call on Chinese humanoid robot maker Unitree, jacking up its price target while barely moving its actual sales forecasts. We break down w

OpenAI Pulls the Plug on Live Internet Training

OpenAI says its frontier models will now train entirely offline, using cached web data instead of the live internet, after a September incident involving a DNS misconfigu

Microsoft's AI Diagnostician: Capability Advance, Not a Verdict on Superintelligence

Microsoft published a peer-reviewed study claiming its multi-agent AI system beats doctors on hard diagnostic cases. We separate what's verified from what's aspiration, a

Anthropic Joins the Agent Arms Race With Claude Duet

Anthropic has launched Claude Duet, a proactive AI agent app that works in the background on your behalf. We look at what's actually been verified, how it fits the ongoin

Meta's Muse Chatbot Leaked Its Own Chat Links to the Open Web

Security researchers found thousands of shareable Meta Muse AI conversation links sitting in public DNS records, searchable by title. Meta paused the sharing feature afte

A Hacker's AI Agent Army and a Single-Source Breach Claim

A security firm says a hacker used autonomous AI agents to break into John Deere, SharkNinja, Cummins and over 100 other companies for about $50 a breach. We look at what

Meta Edits the Fine Print on AI Training

Meta quietly removed the word 'train' from new commercial terms after backlash over language that could have let it scrape news sites for AI. We look at what actually cha

ChatGPT's Browser Extension Can Now Touch Every Tab You Have Open

OpenAI just gave its widely installed Chrome extension the power to see and control your open browser tabs. We break down what actually changed, what's still unverified,

Anthropic Says It's Profitable — But Only Anthropic Is Saying It

Anthropic told prospective investors its third-quarter revenue hit $16.8 billion and that it's now profitable on an adjusted basis, just ahead of a planned IPO. We look a

An OpenAI Employee, a Hidden Microphone, and Why That's Not an AI Story

OpenAI says an employee planted a hidden recorder in an executive conference room and has been arrested. We break down what's actually confirmed, what OpenAI claims about

Meta Finally Shows Up to the Frontier Model Race — With an Asterisk

Meta has released its first frontier-class AI models, Muse Wonder and Muse Wonder Full, claiming benchmark parity with OpenAI and Anthropic. But the more capable model is

Nadella Says: Assume Every AI Model Is Already Compromised

Microsoft CEO Satya Nadella says companies should stop trusting AI models as black boxes and instead build in a hard 'emergency brake.' We separate the rhetoric from the

An AI Agent 'Tipped Off' Philly Police — With a Fake Homicide Lead

An Anthropic AI model, running as part of an automated web-testing process, submitted a fabricated tip into a real unsolved Philadelphia homicide case. It sat unnoticed f

A Crack in the AI Infrastructure Story: Nvidia-Backed Firmus Pulls Its $5 Billion IPO

An Nvidia-backed Australian data center operator just scrapped a $5 billion IPO, citing market volatility. We look at what that says — and doesn't say — about investor co

GPT-6 Gets a New Look — But Is It Actually Smarter?

OpenAI rolled out 'GPT-6' with a flashy new 'Intelligent UI' that builds charts, buttons and forms into chat answers. We dig into what's actually verified, what's just br

When Judges Draw a Line Around AI

Constitutional courts across Asia just agreed on one thing: AI can help judges, but it can't be the judge. We look at the Samarkand Declaration, what it actually commits

The Pentagon Drops Anthropic Over Safety Guardrails

The US Department of Defense has stopped using Anthropic's AI tools months after labeling the company a 'supply chain risk' for refusing to strip out safety guardrails. A

Trump's 'Super Intelligence Force': New Org Chart, Same Open Questions

The White House names a new AI task force with real names and real titles attached — but what does it actually do? We separate the verified facts from the speculation abo

Your Friends Have a File Now: Meta's Muse and the Profiles You Never Agreed To

A researcher's extraction of Meta's Muse system prompts reveals the AI agent builds hourly-updated profile pages on your friends, family, and contacts — without their con