AI safety vs. openness: Amodei on open models and China’s rise
Anthropic’s CEO argues open-weight AI models are valuable but warns against unchecked Chinese AI progress. Here’s why the debate matters.
All news
967 articles — page 9 of 97.
Anti-Data Center Clapping Lands Activist in Custody
A public meeting turns tense as an activist's applause sparks a surprising arrest. What does this say about local tech tensions?
AI-Powered Espionage Hits Thai Finance Ministry
Autonomous AI tool Hermes enabled attackers to bypass controls and steal data from Thailand’s Ministry of Finance in an unrestricted "YOLO mode" operation.
From Half-Baked Code to Production-Ready Chat in Five Days
A developer turned an abandoned side project into a secure, real-time chat app in a single week, showing how small, consistent efforts can finish what once seemed impossible.
AgentENV open-sources scalable sandboxes for AI agents
Moonshot AI and kvcache-ai release AgentENV, a distributed sandbox platform built on Firecracker microVMs to power agentic reinforcement learning at scale.
Google’s AI search overtakes traditional results in nearly half of queries
New data reveals AI Overviews now dominate 43% of Google searches, reshaping how users find information online.
The surprising upside of switching from 360Hz to ultrawide
High-refresh gaming monitors are praised for competitive play, but one user found more value in a wider ultrawide screen—here’s why.
Coca-Cola admits data breach after Fairlife ransomware attack
The beverage giant confirms hackers stole data from its dairy unit Fairlife during a recent ransomware incident, raising fresh questions about supply-chain security.
GitHub Copilot app redefines AI-driven coding workflows
GitHub's new Copilot app introduces a multi-session workspace for AI-assisted development, shifting from chat-only interactions to a more natural, project-focused workflow.
METR’s new AI cost metric: when agents outrun human expenses
A research group introduces a dollar-based "expenditure horizon" to compare AI agents with human labor, but early tests show limits and blind spots.