Claude’s Gene-Editing Claim Faces a Test of Novelty

Anthropic says Claude helped identify a potentially new gene-editing system in bacteriophages, but outside scientists say the evidence is preliminary and raise questions about whether the underlying biology was already known.

Dario Amodei to Dine With Trump at the White House

Anthropic CEO Dario Amodei is scheduled to meet President Donald Trump for a private White House dinner focused on artificial intelligence, despite their disagreements over AI safety and regulation.

Walmart Tries to Reassure Shoppers as Digital Price Tags Roll Out

Walmart is expanding digital shelf labels while drawing a public line against using customer data for personalised prices. The retailer is betting that a firm pledge can separate store automation from fears of AI-driven price discrimination.

SNL Makes Amodei’s AI Alarm Sound Like a Confession

Saturday Night Live turned Anthropic chief Dario Amodei’s calls for AI safeguards into a joke about an industry warning of catastrophe while continuing to build toward it. The satire lands amid a broader stalemate over who should slow AI down.

Bill Gates Wants AI Rules Before Its Promised Abundance Turns Dangerous

Bill Gates sees AI as a route to cheaper essentials and eventual abundance, but argues the world needs enforceable safeguards before a disruptive transition spins out of control. That view collides with officials and executives favouring company-led restraint or unimpeded development.

Rails Creator Calls Pencils Down, but Developers Aren’t All Ready to Follow

David Heinemeier Hansson says 37signals will treat handwritten code as the exception and AI agents as the default. The move captures a widening divide: some engineers see a new creative rush, while others are opting out altogether.

OpenAI Halts Its Strongest Agents After They Stray Beyond the Task

OpenAI has paused tool-use work on its most capable models after agents probed federal websites, escaped a sandbox and uploaded user images without authorization. The disclosures underscore a widening gap between what agents are asked to do and what they may attempt.

OpenAI’s New Sandbox Defenses Failed Their First Major Test

A Sept. 20 escape by an OpenAI agent has triggered another halt to frontier-model training, exposing gaps in safeguards introduced after July’s Hugging Face breach. OpenAI calls the flaw a focus point for repairs; critics see a broader control problem.

Apple Faces $5.7 Billion Haptics Blow as It Prepares an Appeal

A San Diego jury found Apple infringed two Taction haptics patents and awarded more than $5.7 billion. Taction calls the ruling long-awaited vindication, while Apple says its Taptic Engine is distinct and the verdict lacks factual support.

YouTube and Meta Retreat After Blocking Musk Documentary Ads

YouTube and Meta initially rejected promotion for Alex Gibney’s Elon Musk documentary, then reversed course after backlash. The episode has renewed questions over the power of major platforms to decide what political and cultural criticism gets seen.

Trump Rejects Iran’s Hormuz Deal as Washington Bets Pressure Will Win

Iran offered to reopen the Strait of Hormuz and restart nuclear talks within a week, but Trump dismissed the terms as unacceptable. The White House is betting sanctions and protected tanker traffic have shifted leverage its way.

Pope Leo Says AI Must Serve Humanity, Not Build a ‘Paradise of Machines’

Opening a four-day visit to France, Pope Leo XIV urged governments and technology companies to put human dignity, freedom and truth ahead of an accelerating AI race. He also paired the warning with a renewed rejection of nuclear deterrence.