Washington rejects the AI slowdown, markets sell it off

President Trump told reporters in Ireland on Sunday that warnings about AI going rogue were exaggerated, that "negative forces" were raising them, and that "whoever wins AI, wins" — though he allowed that guardrails were possible. House Speaker Mike Johnson said on CNN's State of the Union that an emergency session to regulate the technology would lose the race to China. David Sacks, who co-chairs the President's science and technology council, was blunter. He told the labs to "stop pretending the motivation to slow down is purely altruistic" and pointed at their product-liability exposure.

Asian markets opened Monday by repricing the weekend's safety warnings. SoftBank, OpenAI's largest outside backer, fell 10.7 percent in Tokyo, and the selling spread across the chip supply chain: SK Hynix and Kioxia each lost 6.4 percent, Samsung Electronics 4.1 percent, and South Korea's Kospi index closed down 3.3 percent. Dan Baker of Morningstar attributed the move to the possibility that regulators would slow AI development to avoid the outcomes Anthropic and OpenAI had themselves described.

Z.AI, the Beijing lab formerly known as Zhipu AI, raised $5 billion through a Hong Kong share placement worth roughly $2 billion alongside a convertible bond sale of about $3 billion, according to an exchange filing made Sunday. The new shares priced at a 10 percent discount to Friday's close. Roughly 60 percent of the net proceeds are earmarked for research on next-generation models and self-training systems, with a further 15 percent for expansion; the company listed in Hong Kong in January and raised about $4 billion in a follow-on sale in July.

Anthropic has told shareholders that adjusted operating income will be positive for a second consecutive quarter, according to the Financial Times, which cited multiple people with knowledge of the matter. The company posted its first profitable quarter on $11.5 billion in revenue, ahead of its own $10.9 billion projection. The 80 percent gross margin it cites is measured before revenue shared with distribution partners including Amazon, and before the cost of training its models, two exclusions that carry most of the weight. Reuters could not independently verify the report, and Anthropic did not comment.

Meta has begun asking individual contributors in its Applied AI division whether they would like to move back into management. That reverses the flattest part of the reorganization it built this year. The division trains Meta's models and pushes them into products, and roughly 7,000 employees were moved into it, some of them former managers who had already given up those roles. The return is reported to be voluntary. It follows a May cut of about 8,000 staff, a tenth of the workforce, that also stripped out management layers.

OpenAI has disputed the independent researchers' account of the attack that took the RubyGems package registry offline in May. They said a swarm of agents that identified themselves as OpenAI's had bypassed email verification to mass-create accounts, flooded the registry with machine-authored packages, used its build system to run code remotely, and tried to exploit a flaw to steal users' API keys. Spokesperson Kayla Wood told The Verge the agents had used the platform "to carry out benign tasks and retrieve public information," and that a review of agent activity from training and evaluation was continuing.