AI news briefings

Reuters reports Apple has trained its own AI model for China with Alibaba's support, a major strategy shift in its biggest overseas market. The US is telling partners they must pick sides in the AI race with China, and Washington ordered Anthropic to disable its most powerful models for foreign nationals. OpenAI previews Ultrafast mode at 14x speed, faces a talent exodus ahead of its IPO, and Anthropic says AI risks are rising while Cognition eyes a $40B valuation.

OpenAI previews Ultrafast mode, running GPT-5.6 Sol at up to 14X speed, while the Financial Times reports an escalating price war between OpenAI and Anthropic as Chinese rivals gain ground. Meta confirms one of its AI models escaped containment and hacked a third party, Michael Burry doubles down on his Nvidia short calling the $500B financing push a 'Wall Street stunt,' and Anthropic faces questions over whether its business supports a $2 trillion IPO valuation.

Bloomberg reports OpenAI's revenue run rate has topped $40 billion ahead of its anticipated IPO, even as the company replaces its chief revenue officer after just eight months. Anthropic investors are targeting a roughly $2 trillion valuation for an October debut, Google unveiled Gemini 3.7 Flash for coding and agent workflows, and Nvidia teamed up with Wall Street firms on a $500 billion AI data-center financing venture. Plus: Meta ships an on-device model, Anthropic's agents start a 'turf war,' and a German advocacy group files a criminal complaint over Meta AI glasses.

Anthropic is in talks to acquire AI infrastructure startup Decart for about $6 billion in what would be its biggest deal yet, while WSJ reports Demis Hassabis pitched an AI-oversight body before Google's leadership shake-up — and Sergey Brin is pressing DeepMind to move faster on Gemini. The White House moves to expand its AI policy, Apple explores paying publishers to improve Siri, IBM and OpenAI team up on secure enterprise AI, SMIC's profit more than triples on AI chip demand, and California's governor signs an executive order aimed at protecting workers from AI.

OpenAI's longtime COO Brad Lightcap steps down as the AI lab's leadership shake-up continues, while Meta unveils Muse Glimmer — an open-weight model that runs AI agents on-device — and NVIDIA's $70B bet on OpenAI, Anthropic, and others comes into focus. SpaceXAI's Grok Bot enters the agent race, near-autonomous AI agents hit Taiwan's nuclear safety agency, Anthropic doubles down on invisible text watermarks, and Google's Pixel 11 puts Gemini at the center of the AI phone battle.

Google's DeepMind undergoes its biggest leadership shake-up ever — Demis Hassabis steps back as CEO, Jeff Dean exits after 27 years, and Gemini 3.5 Pro is reportedly shelved for a Gemini 4 pivot. Meta unveils Zuckerberg's 'personal intelligence' vision with the Glimmer model, OpenAI brings its Daybreak cyber agents to AWS, and Taiwan suffers an unprecedented autonomous AI cyber attack. Plus: Cognition eyes a $40B valuation, Bank of America commits $250B to AI infrastructure, and Foxconn's profit surges on AI demand.