Japanese AI-linked stocks drop in sync with Wall Street semiconductors, forcing investors to reprice near-term AI hardware demand risk.

South Korea says it will train a security-first sovereign model, after US moves that limited access to Anthropic’s Mythos.

The China EV maker says it will scale Iron humanoid robot production to enable a global launch, racing Tesla.

The “roughly 300” figure signals AI adoption is accelerating fast, not slowing down.

After GPT-5.6 launched July 9, reports of unauthorized deletions exposed how permission defaults and agent sandboxing collide.

The regulator says HP colluded with channel partners to raise bid costs and squeeze resellers, triggering a major enforcement signal.

Creators get AI-powered personas, and Epic gets a new platform layer that raises quality, rights, and safety stakes.

The hack halts operations, so executives get a real-world stress test of continuity, compliance, and supply-chain risk.

A new browser integration lets Claude sign into websites using stored 1Password credentials without sending passwords to the AI model.

The streamer is scaling AI production workflows, citing faster output and lower costs, plus real examples.

A second-quarter earnings letter lays out how generative AI spreads from concept to post-production, and why it changes the workflow.

A one-hour experiment exposes why open-weight does not mean inspectable, and why security teams cannot treat models like code.