Agents & Coding
Meta's Muse has brought new attention to AI agent harnesses, software that gives models memory, tool access and the ability to complete tasks. Analysts say Chinese companies including Tencent, Alibaba and ByteDance have strong starting points, but fragmented platforms, subscription habits and computing costs could slow their progress.
Policy & Safety
US President Donald Trump has rejected calls for joint work with China on artificial intelligence risks, saying cooperation could weaken America's lead. His position came as Google, Nvidia, Anthropic, OpenAI, Meta Platforms and SpaceXAI signed a voluntary White House accord promising internal controls, outside audits and regular discussions on safer AI practices.
Policy & Safety
Meta says Muse is restricted to adults and requires age checks, but its cuddly Jolly mascot has prompted criticism from youth advocates and concern from researchers who say cute design can lower users' guard around privacy, trust and the role of an AI agent.
Policy & Safety
In an NBC News interview, Meta chief Mark Zuckerberg rejected proposals for industrywide coordination to slow artificial intelligence development. He said commercial incentives could encourage companies to build AI safely, while dismissing extinction warnings as doom-focused rhetoric and describing Meta's Sentinel agent as a safeguard for its Muse system.
Policy & Safety
Meta released a hotfix for a 0-day flaw in Muse, its macOS AI assistant, after security researcher Patrick Wardle found that attackers could redirect speech processing, obtain an authentication token and use Muse's broad permissions to control accounts and perform actions on a victim's Mac.
