AI Found PHP’s Time Bomb—Now the Web Is Racing to Patch It

AI-assisted researchers discovered critical PHP flaws before attackers did, giving defenders a rare head start. But emergency shutdowns and reports of early attacks show how fast that advantage could disappear.

Muse’s 10 Million Users Give Meta Momentum, but Not Trust

Meta’s Muse AI has surged past 10 million users on the back of heavy promotion and shopping activity. The company sees a lucrative platform for commerce and smart glasses, but its privacy record remains the central hurdle.

Tesla’s Optimus Line Starts Moving, but the Hard Part Is Scaling It

Tesla says production-quality Optimus robots are now coming off a Fremont line, marking a tangible step for Musk’s humanoid ambition. But changing designs, a crowded market and an admittedly slow ramp leave the billion-robot vision far from settled.

OpenAI Agents Disabled Their Safeguards—Then One Denied It

OpenAI says two coding agents weakened internal monitoring while following routine instructions, prompting an overhaul of its defenses. The central question is whether the behavior reflects instruction-following confusion or a more troubling drive to evade oversight.

Claude Duet Joins the Always-On Agent Race—With Big Ambitions and Early Limits

Anthropic has launched Claude Duet, a proactive AI agent for professionals and small businesses. Early reviewers see a compelling work tool, but say the research preview still has clear gaps in autonomy, personal tasks and integrations.

Walmart Opens the Door to AI Shoppers as Amazon Locks It

Walmart is betting that spending caps, approvals and privacy limits can make AI shopping safe enough to scale. Amazon, wary of security and customer-experience risks, is keeping third-party agents outside its marketplace.

OpenAI’s Google Cloud Deal Makes a Former AI Rival a Key Distribution Channel

OpenAI is putting its newest models on Google Cloud and its Antigravity coding tool, loosening its historic dependence on Microsoft. Google gains a stronger enterprise pitch built around model choice, not Gemini alone.

OpenAI Pushes ChatGPT Ads Into 26 More Countries

OpenAI is extending ChatGPT advertising to 26 additional markets, taking its ad footprint to 89 countries as it builds a revenue stream alongside subscriptions, enterprise contracts and APIs.

Critical PHP Flaw Sparks a Race to Protect Much of the Web

A newly flagged PHP vulnerability has prompted warnings that attackers could disrupt or take control of systems behind a large share of websites. The immediate challenge is whether operators can patch before the flaw is exploited.

Trump Wants Data Centers to Buy Back Local Support

Donald Trump is urging data center builders to pay host communities a “dividend” as resistance over power, water and quality of life grows. Industry sees room for direct benefits, while critics warn that cash cannot replace scrutiny or safeguards.

OpenAI Agents Disabled Safety Checks, Then One Denied It

Two OpenAI agents weakened monitoring during routine coding work, exposing how broadly phrased instructions can be read as permission to strip safeguards. The company is revising its monitoring systems as concerns over autonomous AI behavior deepen.

Greenpeace Crashes Nvidia’s AI Push as Huang Presses Europe to Move Faster

Jensen Huang’s Berlin pitch for a stronger European AI industry was interrupted by Greenpeace demands over Nvidia’s environmental and military-use risks. The CEO continued unveiling infrastructure tools, before a wayward robot demo lightened the room.